Stigma and Other Determinants of Participation in TANF and Medicaid [PDF]

Association for Public Policy Analysis and Management

Stigma and Other Determinants of Participation in TANF and Medicaid Author(s): Jennifer Stuber and Karl Kronebusch Source: Journal of Policy Analysis and Management, Vol. 23, No. 3 (Summer, 2004), pp. 509-530 Published by: Wiley on behalf of Association for Public Policy Analysis and Management Stable URL: http://www.jstor.org/stable/3326264 Accessed: 10-04-2017 02:21 UTC REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/3326264?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms

Association for Public Policy Analysis and Management, Wiley are collaborating with JSTOR to digitize, preserve and extend access to Journal of Policy Analysis and Management

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Stigma and Other Jennifer Stuber Determinants of Participation Karl Kron

in TANF and Medicaid

Abstract

We developed a conceptual framework to examine the association between stigma, enrollment barriers (e.g., difficult application), knowledge, state policy, and participation in the Temporary Assistance to Needy Families (TANF) and adult Med-

icaid programs. Survey data from 901 community health center patients, who were potential and actual participants in these programs, indicated that while

images of the Medicaid program and its recipients were generally positive, stigma associated with welfare stereotypes reduced both TANF and Medicaid enrollment. Expectations of poor treatment when applying for Medicaid, enrollment barriers, and misinformation about program rules were also associated with reduced Medicaid enrollment. States that enacted strict welfare reform policies were potentially decreasing TANF participation, while states with more simplified and generous programs were potentially increasing Medicaid participation. The results suggest that the image of the adult Medicaid program remains tied to perceptions about welfare and provides guidance to policymakers about how to improve participation rates. ? 2004 by the Association for Public Policy Analysis and Management. INTRODUCTION

Low-income families in the United States frequently do not participate in meanstested government programs for which they are eligible. Estimated take-up rates

range from 40 to 70 percent for programs providing cash assistance (Blank and Rug-

gles, 1996; Moffitt, 1987) and in-kind benefits, such as Medicaid (Seldin, Banthin, and Cohen, 1998) or food stamps (Blank and Ruggles, 1996). The implementation

of welfare reform appears to have made this problem worse. Welfare take-up rates fell from 84 percent in 1995 to 56 percent in 1998 (Zedlewski, 2002), and there have

been concurrent declines in Medicaid and food stamp enrollment for both adults and children (Klein and Fish-Parcham, 1999; Kronebusch, 2001; Zedlewski and

Brauner, 1999), despite the expectation that eligibility for Medicaid and food stamps had been preserved by federal law. Understanding why eligible families do not enroll in these programs is critical to devising effective outreach, enrollment, and retention strategies.

Researchers and policymakers have long believed that stigma and other barriers may account for the nonuse of program benefits among persons who are eligible.

Manuscript received September 2002; review complete March 2003; revision complete April 2003; revision review complete September 2003; accepted December 2003

Journal of Policy Analysis and Management, Vol. 23, No. 3, 509-530 (2004) ? 2004 by the Association for Public Policy Analysis and Management Published by Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/pam.20024

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

510 / Determinants of Participation in TANF and Medicaid

Three different approaches have been used to examine the relationship between

stigma and participation. First, surveys of low-income families have assessed perceptions of programs and reasons for nonparticipation (Perry et al., 2000; Shrup-

tine, Grant, and McKenzie, 1998). However, these studies are often limited to

descriptive comparisons and do not control for confounding factors that may also affect program enrollment.

The second approach develops models of formal decisionmaking, typically in a utility-maximizing framework. Because individuals are assumed to be rational,

non-participation in programs that provide benefits must be due to a potential participant's "distaste" for the program, which is then used as the operational definition of stigma. For example, Moffitt (1983) models welfare participation and labor supply decisions, with parameter estimates derived from the differences in enrollment and hours worked for program participants and nonparticipants. Stigma here is not directly measured, but is assumed to be present because individual responses to the availability of resources appear to differ by source-whether from wages, non-wage income, or program benefits. The main drawback to this approach is that without directly measuring stigma, it is difficult to draw inferences about the relative impacts of stigma and other barriers on program participation. Instead, the estimated "stigma" effect results from all of the potential reasons for nonparticipation, as well as potentially unmeasured differences between participants and nonparticipants. A third strategy is to use proxy variables that indirectly measure the presence of stigma. For example, Currie and Grogger (2000) estimated the effect of stigma by whether a state had implemented electronic benefit transfers for the food stamp program, which, because the use of electronic debit cards is potentially less visible than food stamps themselves, may result in less stigmatized experiences. The value of any proxy variable approach depends on how closely related these variables are to the underlying beliefs or perceptions. Because stigma is more multi-faceted than can be readily captured by an imperfectly measured proxy variable, for this study we use data from a survey that was designed to measure directly, from the viewpoint of potential and actual recipients,

stigma associated with means-tested programs. Utilizing concepts drawn from social psychology and sociology permits a more nuanced view of stigma than has been possible in previous studies. Because the role of stigma, a lack of knowledge, and enrollment barriers may vary by program, the empirical analysis examines two related programs-cash welfare benefits and adult Medicaid.

CONCEPTUAL FRAMEWORK

A common analytic starting point is to assume that potential program participants "rationally" weigh the subjective costs and benefits of participation. They participate when the perceived value of the benefit exceeds the cost of participating (Barr and Hall, 1981; Willis, 1980). There are clear benefits derived from participation in means-tested government programs, particularly if the programs are perceived to meet people's needs effectively. However, both monetary and non-monetary costs are also incurred, including stigma and other barriers to enrollment. Stigma

A stigma is "an attribute which is deeply discrediting" in a given society (Goffman, 1963, p. 3). We distinguish two forms of stigma-identity and treatment stigma-

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Determinants of Participation in TANF and Medicaid / 511

that may arise when claiming assistance from government programs. Identity stigma is related to concerns about being labeled with negative stereotypes associated with recipients of means-tested programs. These stereotypes correspond with broadly held American views that equate social status with personal achievement (Kerbo, 1976). Consistent with this individualistic belief system, being poor is considered deviant behavior because pulling oneself out of poverty is considered both possible and praiseworthy. Those who do not succeed in escaping poverty are viewed as meriting this fate (Rainwater, 1982). Negative stereotypes of public assistance recipients also arise from social beliefs that recipients of means-tested government programs are undeserving (Cook and Barrett, 1992; Gilens, 1999). As a result, people on public assistance have been labeled lazy, lacking in ambition, shiftless, dishonest, aggressive seekers of unearned rewards, morally weak, and bad parents (Hochschild, 1996; Klugel and Smith, 1985). These stereotypes are conveyed through a variety of social processes, including the news media (Gilens, 1999; Iyengar, 1990) and social pressure. Exposed to these stereotypes and their possible isolating effects, program recipients may internalize them, adopting negative selfcharacterizations and completing the internal stigmatization process (Horan and Austin, 1975). Fear of being labeled in this fashion may deter potential recipients from participating in government programs. In addition, the act of seeking public benefits exposes individuals to a potentially hostile environment. While the first form of stigma captures how individuals view themselves, the second form of stigma-treatment stigma-captures how others view and treat recipients. Research has shown that once stigmatized, individuals are labeled as responsible for their fate, and are met with anger and limited willingness to help (Weiner, Perry, and Magnusson, 1988). This stigmatized treatment arises in a wide range of social contexts. Consistently, recipients describe their experiences on public assistance as unpleasant and negative (Auleta, 1982; Goodban,

1985; Piven and Cloward, 1993). Treatment stigma may inhibit participation

because potential recipients fear they will be treated poorly while participating in public assistance. Concerns about identity and hostile treatment in the welfare program are typically considered a general attribute of all means-tested programs. Generally, when researchers and policymakers discuss stigma in the Medicaid program, they presume that this stigma is derived from the program's association with welfare (Mann, 1999). The historic and continued administrative connections between the two programs may mean that the potential consequences of stigma resulting from participation in Medicaid continue to be undifferentiable from those associated

with welfare.

However, stigma may also differ across programs. Altruistic support for program benefits may be systematically related to how taxpayers and donors expect the benefits will be used (Nichols and Zeckhauser, 1982). Medicaid may be perceived as less susceptible to abuse than welfare because recipients receive medical care benefits, which are presumed to be appropriately limited to defined needs, rather than cash benefits, which might be spent inappropriately (Cook and Barrett, 1992). Perceptions of abuse can contribute both to the stereotyping of people who are enrolled in welfare and Medicaid and to the poor treatment of people applying for the benefits. It is likely that Medicaid recipients (who are not on welfare) will be perceived as more deserving of the benefit because it is widely recognized that many people who are working do not have access to health insurance (Jacobs and Shapiro, 1996). There may also be concerns about poor treat-

ment in the Medicaid program that do not apply to cash assistance because

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

512 / Determinants of Participation in TANF and Medicaid

recipients using Medicaid must reveal their stigmatized status to health care

providers and their office staff. Other Individual Determinants Perceived Enrollment Barriers

A number of other costs to obtaining benefits may also exist. The application

process for means-tested programs has been characterized by many as labor-intensive, requiring an in-person visit to an often inconveniently located office during limited office hours. Application forms can be confusing and sometimes must be accompanied by multiple forms of supporting documentation, each piece requiring verification. Transportation can be a problem for low-income families, especially when applicants are expected to make several trips to the welfare office, both during the initial application process as well as to respond to eligibility re-determinations, which may require additional documentation. Translators may not be avail-

able for non-English-speaking applicants and application forms may not be

available in foreign languages. A Lack of Knowledge

In addition, eligible persons may not participate in means-tested programs because they do not understand program rules or how to go about applying for benefits (Kenney, Haley, and Dubay, 2001). Misinformation may negatively affect participation in several ways. Eligibility requirements for TANF and Medicaid are complex, vary widely among states, and have changed significantly over the last 15 years. Under federal mandates of the 1980s and 1990s, Medicaid eligibility was extended beyond those receiving welfare benefits to cover low-income pregnant women, children, and former welfare recipients who are no longer receiving cash benefits. Still, many families believe they have to be on welfare to get Medicaid, think eligibility

for Medicaid applies only to women when they are pregnant, do not know how

much money they can earn and still qualify for Medicaid, and so forth (Andrulis, Bauer, and Hopkins, 2001). Eligibility requirements may be particularly confusing to immigrants because of eligibility changes in the 1996 welfare reform law and subsequent legislation, as well as because they may fear that seeking government help will affect their immigration status or jeopardize their opportunity to become a citizen (Fix and Zimmerman, 1998; Maloy et al., 2000). The implementation of the

1996 federal welfare legislation may have generated additional confusion about

Medicaid eligibility, especially about whether the work requirements, time limits, and sanctions of welfare reform affect Medicaid eligibility (Perry et al., 2000). Need

Needier individuals may be more likely to participate in TANF and Medicaid because they place more value on the benefit than do people who are less needy. Individuals in greater need include those who have lower income, fewer employment opportunities, poor health, and more children. For example, those in poor

health may be unable to work (or to work long periods), and so the cash benefit may be essential. They may also be more likely to seek medical services, and so they will value having health insurance more than those who are in good health.

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Determinants of Participation in TANF and Medicaid / 513

State Policy

Individuals make participation decisions within a broader context of program design and implementation. States have considerable discretion in designing and implementing TANF and adult Medicaid. For example, states set eligibility standards within defined federal limits, implement welfare reform, can improve the generosity of the Medicaid benefit, and can simplify Medicaid enrollment proce-

dures.

Eligibility Stondards

More generous eligibility standards expand the pool of potentially eligible recipi-

ents. The impact of eligibility standards on program participation can be measured in several different ways. Using the eligibility standards and earned income disregards for the relevant states, we simulated welfare and Medicaid eligibility for each

respondent based on reported household income, the number of children in the household, and the number of adults in the household. Second, the state-set gross income cutoffs for welfare and Medicaid were also included, to test for the possi-

bility that broader eligibility standards improve take-up, that is, increasing the likelihood that eligible recipients will pursue enrollment, as well as expanding the pool

of eligibles. Third, we included an interaction term based upon simulated individual eligibility and reported income to test whether enrollment changes as income

rises to levels close to the eligibility cut-off. Because of the way benefits are computed, potential applicants whose income is just below the eligibility cutoff may be eligible for only a few dollars in cash benefits, although they still receive the full value of the Medicaid benefit. This serves as an incentive for participation in Med-

icaid for those close to the eligibility standard, even if the cash benefit is small (Yelowitz, 1995).

Welfare Policy

TANF block grants created under the 1996 federal welfare law gave states discretion in implementing their welfare programs and, as a result, state policies with respect

to time limits, sanctions, mandatory applicant job searches, and welfare diversion programs vary substantially (State Policy Documentation Project, 2000). First, the federal welfare reform law required states to establish TANF time limits of 24

months of continuous cash assistance or 60 total months, but states were given the

option to enact stricter time limits, and a number of states have done so. Second, failure to meet work participation requirements can result in financial sanctions,

which in some states includes full termination of TANF benefits at the first instance

when a client fails to comply. Third, while most TANF programs stress moving recipients into employment quickly, some states require applicants to have completed a job search as a pre-condition of applying for TANF. Fourth, in keeping with this work-first philosophy, many states have adopted welfare diversion programs which provide potential TANF applicants with one-time assistance, such as offering clients a small lump sum payment if they agree to stay off cash assistance for one

year (Maloy et al., 1998).

While state welfare policies directly affect TANF participation, they may also have

had an indirect impact on Medicaid enrollment (Chavkin, Romero, and Wise,

2000). When enacting the 1996 welfare reform law, federal policymakers did not

include Medicaid in the welfare-related block grants to states. Instead, they explic-

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

514 / Determinants of Participation in TANF and Medicaid

itly preserved Medicaid eligibility for adults under the 1996 welfare rules, and

allowed states to expand Medicaid access. Nevertheless, the implementation of welfare reform may affect Medicaid. For example, families applying for welfare diversion payments or who must complete a mandatory job search prior to applying for

TANF are supposed to receive Medicaid in the interim. However, concerns have

been raised that this may not always occur (Maloy et al., 1998). In addition, appropriate safeguards may not be in place to guard against the erroneous termination of Medicaid cases as recipients reach the federal and state set time-limits (Dion and Pavetti, 2000). Medicoid Generosity State policies may also affect applicant perceptions about the value of the Medicaid program. Higher provider reimbursement rates have been shown to increase access

to providers (Cohen and Cunningham, 1995; Long, Settle, and Stuart, 1986), and expanded coverage of Medicaid optional services may increase the value of the

Medicaid program for recipients. Moreover, the value of the Medicaid benefit has been shown to increase the likelihood that a single mother will apply for and stay on welfare (Blank, 1989; Moffitt and Wolfe, 1993), and so, the generosity of the Medicaid program may also affect TANF enrollment. Medicaid Enrollment Simplification

Advocates and policymakers have called on states to simplify their Medicaid appli-

cation processes to increase enrollment, and many states have simplified enrollment for children's Medicaid (Cohen Ross and Cox, 2000). While simplification strategies have not been applied as extensively to the adult Medicaid program, states have begun to modify these procedures. As well as adopting Medicaid-only

and mail-in applications, some have eliminated face-to-face interviews and dropped resource tests for its adult Medicaid program (Maloy et al., 2002). METHODS

Study Design and Data Collection Patients were interviewed at 23 community health centers in 10 states (California,

Colorado, Idaho, Massachusetts, Michigan, Missouri, Pennsylvania, South Car-

olina, Texas, West Virginia) and the District of Columbia between April and December 1999. The states were selected to ensure variation with respect to geography, population size, political ideology, and culture (Elazar, 1984; Erikson, Wright, and Mclver, 1993). Leadership at the National Association of Community Health Centers identified community health centers within each state that would reflect the diversity of the state including health centers of varied size, from both urban and rural areas. From this list, we selected an urban and a rural health center at random.

In all, 1686 people were approached to participate in the study; 281 refused or

were unable to complete the entire interview, resulting in a final sample of 1405 completed surveys, or 83.3 percent of the eligible sample. The vast majority of those who did not complete the survey were unable to do so because of time constraints. Respondents were chosen by a simple sampling strategy. Patients were interviewed

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Determinants of Participation in TANF and Medicaid / 515

throughout the day beginning with the first person to arrive at the clinic for a

patient visit. The next person to sign in after the interviewer completed her previous interview was selected. Willing participants completed the survey while waiting to be seen by a provider or at the conclusion of their medical visit. Research staff administered the surveys in person with interviews lasting between 20 and 30 min-

utes. Participants were offered $10 compensation and were given the option to be interviewed in Spanish (16 percent) or English (84 percent). Community health centers serve a low-income population without regard to their insurance coverage, and typically serve a mix of Medicaid recipients and those who

are uninsured. Collecting the data at health centers allowed us to easily identify low-income adults who are potential or actual participants in public assistance programs. Because not all community health center patients are potential or actual participants in TANF or Medicaid we restricted our final sample for this analysis to adult respondents with incomes below 300 percent of the federal poverty level (FPL) who have at least one child under the age of 18 in the household. Although adults with current income above the poverty line will generally not be eligible for TANF or Medicaid, the high degree of income fluctuation in the low-income population means that many of these individuals may have been eligible for benefits in the recent past. Respondents were generally representative of community health center patients in these states in terms of their race or ethnicity and insurance status. In these states, 35 percent of community health center patients were white, 26 percent black, 31 percent Hispanic, and 8 percent classified as another race. In this sample, 29 percent of respondents were white, 31 percent black, 34 percent Hispanic, and 6 percent classified as another race. In these states, 40 percent of community health center patients were uninsured and 34 percent were enrolled in Medicaid, whereas, in our sample, 39 percent of community health center patients were uninsured and 34 percent were enrolled in Medicaid (Bureau of Primary Health Care, 1999). Compared with data drawn from the March 2000 Current Population Survey, a nationally representative sample of adults with children under age 18 with family income less than 300 percent of poverty from these 10 states and the District of Columbia, our survey respondents were more likely to be black and to be uninsured or enrolled in Medicaid.

Measures

Multiple items were used to measure each of the constructs discussed above (se Table 1). The constraints of collecting data while respondents were waiting for appointment required these concepts be assessed with fairly short item scale Scales were created using principal components factor analysis, and for each sca Cronbach's alpha statistic was used to assess how well a set of items measured a single unidimensional latent construct. Survey questions measuring stigma, enrollment barriers, and lack of knowled were generally asked of respondents in an indirect way with the assumption th people's responses about others would be informed by views of themselves. Because many of the questions required respondents to ascribe negative characteristics themselves (e.g., laziness, being misinformed), the goal of asking questions in t third person was to get more reliable responses. In addition, not all responden had experience with TANF or Medicaid. The use of indirect questions allowed us measure broad perceptions among all respondents, even if they could not provi answers based on their direct personal experiences.

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Table 1. Variable definitic Variable name

Identity stigma

)ns. Ir

I1

variablecomoonents --ru-

1

.

-%AAA.7mLA

.

.

Vaiahile co

%A.-W

Welfare

1) I worry that being on welfare would make me lazy Responses ranged from 2) People in the country on welfare are lazy agree on a five-point L 3) Women on welfare do not take good care of their kids computed based on ave 4) Welfare makes people work less than if Cronbach's there weren't a alpha = 0.52 welfare system 5) Many people on welfare do not want other people to kno 6) A lot of people in this country don't respect a person on

Medicaid

1) I worry that being on Medicaid would make me lazy Responses ranged from 2) Many people in this country on Medicaid are lazy agree on a five-point L 3) Many people on Medicaid do not want other people to knowcomputed based on ave 4) A lot of people in this country don't respect a person on Cronbach's alpha = 0.5

welfare

Medicaid

Treatment stigma

Applying and 1) The application process for welfare (Medicaid) is participating in welfare humiliating

Comparable items were Medicaid. Responses ra 2) Many people are treated poorly when they apply for welfare strongly agree on a fiv (Medicaid) somewhat or strongly 3) When applying for welfare (Medicaid) you have to answer responses were coded as 0 (1 unfair questions about your personal life created two variables rang 4) When participating in welfare (Medicaid) the rules take two variables were indexes away your personal freedom computed. Treatment by health 1) Doctors do not provide high quality healthcare to people Responses on ranged from str care providers Medicaid agree on a five-point Liker 2) Doctors do not treat people on Medicaid equal to people computed based on averag with private health insurance Cronbach's alpha = 0.67.

Other individual determinants

Enrollment barriers 1) It is not worth the hassle to apply for welfare (Medicaid) Comparable items were ask 2) The application for welfare (Medicaid) is long and Medicaid. Responses ranged complicated strongly agree on a five-p 3) It is hard to get the documents needed to apply for welfare somewhat or strongly agr

(Medicaid) responses were coded as 0 (1 4) The hours when the welfare (Medicaid) office is open arecreated not two variables rang convenient two variables were indexes 5) It is difficult for people to find transportation to the welfare computed.

(Medicaid) office 6) Many people find it hard to find translators when applying for welfare (Medicaid) (Table I continued)

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Table 1. Variable definitions (cont.) ....... r * VIa Variable name General confusion

about eligibility

Misinformation about Medicaid rules

Need

Demographics Race

c

s

*1

Itanae

Variable components

ons_rc to variable c

Comparable items were welfare Medicaid. Responses ran strongly agree on a fiv somewhat or strongly responses were coded a created two variables r two variables were inde computed. Responses were true or false 1) You have to be on welfare to get Medicaid (False) coded as 1, incorrect respon 2) The work requirements for the welfare program apply to summed. In the analysis we people on Medicaid (False) knowledge about Medicaid r 3) Welfare time limits apply to people on Medicaid (False) 4) People can apply for Medicaid at places other than a welfare0-4 and as a binary variable knowledge and 1 less knowle office (True) negative effects of knowled when respondents have high (cut point answered 3 or 4 q not a gradual function of hav Because this variable was an wvas not computed. Coded continuously 1) County unemployment rate Responses of excellent, very 2) Health is excellent, very good, good, fair, poor? as 0 and responses of fair or Coded continuously 3) Number of children in household 1) Many people do not know how or where to apply for (Medicaid) 2) Many people are confused about who is eligible for welfare (Medicaid) 3) Immigrants are afraid to apply for welfare (Medicaid)

White, Hispanic, black and other race

Age

Three dummy variables wer Hispanics to whites, blacks t other races compared to wh Coded continuously

Gender Education

Female = 1, Male = 0

Number of parents

Two parents = 1, single Coded continuously

High school graduate = 1, Le education = 0

Annual income

(Table 1 continued)

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Table 1. Variable definitions (cont.) Variahle name * ax. m&ah

State policy

Simulated eligibility

Allol.

Variahle c-amnfnnnteq VaIAAv , Cva

lauJIC;

Va1

1) Annual household income 2) Number of children and adults in household 3) State eligibility cut-offs

4) State earned income disregards State eligibility cutoffs State eligibility cut-offs

Interaction: Income & I) Simulated eligibility eligibility cutoff 2) Annual household income Welfare reform

1) Cash benefits terminated at first sanction

2) Mandatory applicant job search as a condition of eligibility 3) Presence of a formal diversion program 4) Time-limits that take effect earlier than the federal requirement

Medicaid simplification 1) Applicants can mail in applications 2) Has a Medicaid only application Medicaid generosity

lltl

cnstrlu

Eligibility dete were made by state eligibility

controlling for

household

The state's monthly gro in $$, multiplied by 12 Interaction term betwe

annual household income

The welfare reform va sum of the number of

enacted in the state. Becaus

index, coefficient alpha

The Medicaid simplific

was the sum of the num 3) Elimination of face-to-face interviews policies enacted in the 4) Elimination of the asset test an index, coefficient al 1) Medicaid expenditures above the mean for the sample The Medicaid generosit 2) State offers partial or full dental benefits to adult recipients was the sum of the num enacted in the state.

Regional and local

effects

South Midwest

WV, SC, TX, DC

West

CA, ID, CO MA, PA Health center is located in an urban or a rural area

Northeast Urban

MO, MI

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

South = 1 Midwest = 1 West = 1

Urban = 1, Rural = 0

Determinants of Participation in TANF and Medicaid / 519

Statistical Analysis

The dependent variables in this study were current enrollment in TANF (no = 0, yes = 1) and current enrollment in Medicaid (no = 0, yes = 1). We fit models for each dependent variable using the explanatory variables discussed above, adjusting for individual-level demographic characteristics as well as regional and local fixed

effects. Generalized estimating equations were used to adjust the standard errors of the parameter estimates because of the non-independence of responses of individuals by state and community health centers.

RESULTS

.Table 2 compares the survey responses among those who were enrolled and those who were not enrolled in TANF and Medicaid. In general, those who were not enrolled in either program were significantly more likely to report welfare and Medicaid stereotypes, to perceive poor treatment or enrollment barriers, and to be more confused about eligibility. For example, 66 percent of people not enrolled in welfare

agreed with the statement, "Many people on welfare do not want other people to

know they are on welfare," compared with 52 percent of people who were enrolled in welfare. Comparing the two programs, negative stereotypes associated with peo-

ple on welfare were more than twice as prevalent as Medicaid stereotypes. For

example, 69 percent of respondents not enrolled in welfare said, "A lot of people in this country don't respect a person on welfare," whereas 33 percent of respondents not enrolled in Medicaid said, "A lot of people in this country don't respect a person on Medicaid." Also, bad experiences associated with applying for benefits were

much more common for cash benefits than for Medicaid: 45 percent of those not enrolled in welfare perceived the application process as humiliating versus 24 percent for Medicaid. By contrast, confusion about eligibility was more of a problem

in the Medicaid program: 79 percent of those not enrolled reported this for Medicaid compared to 71 percent for welfare. There were also observable differences with respect to policies in the sampled states: more stringent welfare reform policies, less

expansive eligibility for TANF, and less generous Medicaid programs appeared to reduce enrollment, while simplification of the Medicaid application and more generous Medicaid benefits appeared to increase enrollment. Table 3 presents the mean values of the explanatory variables included in the analysis. Tables 4 and 5 present multivariable models showing factors associated with TANF and adult Medicaid enrollment. In both, the first model specification represents a simplified model limited to measures of need and demographic characteristics. The second model includes the two forms of stigma, enrollment barriers, and a lack of knowledge. The third model looks at the effects of state policy vari-

ables controlling for regional and local fixed effects, but excludes the variables measuring stigma, enrollment barriers, and lack of knowledge. The final column

presents the full specification. The Medicaid models include several additional vari-

ables specific to the Medicaid program, including measures of poor treatment by

health care providers, a lack of knowledge about Medicaid program rules, and Med-

icaid simplification. To capture potential cross-program effects, the Medicaid model includes the measure of identity stigma associated with welfare to assess if stigma resulting from participation in Medicaid stems from the program's historic association to welfare. The welfare model includes the measure of Medicaid program generosity to assess if the value of the Medicaid benefit affects the likelihood of participation in welfare.

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

520 / Determinants of Participation in TANF and Medicaid

Table 2. Bivariate analysis of factors associated with TANF and Medicaid enrollment. Variable TANF Medicaid

Percentage wh answer corre (N = 874) (N = 901) Not

Not

Enrolled enrolled Enrolled enrolled

Identity stigma

(N = 176) (N = 698) (N = 386) (N = 515)

Welfare stigma

Being People Women Welfare

on welfare would make me lazy 22% 34%** 27% 35%** in this country on welfare are lazy 36% 44%* 37% 47%** on welfare don't take good care of their kids 20% 24% 20% 25%* makes people work less than they would if there 52% 63%** 58% 67%*

weren't a welfare system

Many people on welfare do not want other people to know 52% 66%** 53% 71%* A lot of people in this country don't respect those on welfare 63% 69%* 67% 70%

Medicaid stigma

I worry that being on Medicaid makes people like me lazy 8% 12%** Many people in this country on Medicaid are lazy 14% 14% Many people on Medicaid do not want other people to know 25% 37%** A lot of people in this country don't respect those on Medicaid 24% 33%**

Treatment stigma

Poor treatment while applying and participating

The application process is humiliating 37% 45%* 15% 24%** Many people are treated poorly when they apply 47% 46% 19% 30%** When applying you have to answer unfair personal questions 46% 56%** 27% 40%** When participating the rules take away your personal freedom 28% 33% 14% 18%**

Poor treatment by providers

Doctors do not provide quality health care to people on Medicaid 14% 16% Doctors do not treat people on Medicaid equal to people 30% 36%* with private insurance

Other individual determinants

Enrollment barriers

It's not worth the hassle to apply 22% 37%** 24% 30%** The application is long and complicated 48% 56%* 30% 48%** It's hard to get papers needed to apply 39% 37% 25% 34%** The hours you can apply are not convenient 21% 28%* 23% 30%** It's hard to find transportation to apply 32% 38% 32% 36% It's hard to find translators to help apply 38% 45%* 39% 43% Confusion about eligibility

Many people do not know how or where to apply 35% 46%** 48% 61 %* Many people are confused about who is eligible 57% 71%** 64% 79%* Immigrants are afraid to apply 50% 61%** 55% 65%*

Less knowledge about Medicaid rules

You have to be on welfare to get Medicaid 81% 71%?*'* The welfare work requirements apply to people on Medicaid 48% 47% The welfare time limits apply to people on Medicaid 54% 44%** Respondent knows about applying for Medicaid at 47% 45% non-welfare offices

Need

Poor health status 33% 26%** 28% 27% Number of children 40% 27%** 33% 26%*

Unemployment rate is greater than the mean for the samplea 38% 27%** 30% 18% State policy

Welfare reform

State terminates benefits at first sanction 5% 22%** 11% 24%**

State requires a mandatory job search at time of application 55% 63%** 65% 58%** State has a formal cash diversion program 35% 45%** 29% 53%** State reached welfare time limit prior to January 2000 27% 40%** 32% 42%** (Table 2 continued)

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Determinants of Participation in TANF and Medicaid / 521

Table 2. (continued) Variable

TANF

Medicaid

Percentage w answer corr (N 874) (N 901) Not

Not

Enrolled enrolled Enrolled enrolled

(N = 176) (N = 698) (N = 386) (N = 515) State policy (continued) Medicaid simplification

State eliminated in-person interview 89% 80%** State

eliminated

asset

test

53%

31%**

State uses separate Medicaid-only application 75% 59%** State allows families to mail in application 89% 80%** Medicaid generosity

State family adult Medicaid expenditures above the meanh 34% 25%** 29% 25%**

State provides Medicaid coverage for adult dental care 68% 56%** 64% 55%** Eligible for TANF or Family Medicaid 81% 38%** 71% 39% Gross income cutoff is greater than the mean for the sample TANF ($8,975) 72% 43%**

Medicaid

($11,852)

46%

26%**

a The mean unemployment rate for the The mean adult Medicaid expenditures f ** Significant at the p < 0.05 level.

*

Significant

The

at

results

the

in

p

<

0.10

Table

level.

4

illus

being labeled by welfare tion. At the same time, p stigma) did not have a sig ment

barriers

and

confusio

ticipation. Policy reforms reduced TANF enrollment

enrollment.

hood of TANF enrollment.

All

of

the

vari

To assess the magnitude of these associations, we calculated marginal probability changes by varying each explanatory variable by one standard deviation above and below the mean, while holding the other variables constant at the mean for the sample. The mean probability of being enrolled in TANF was 0.10 in this sample. Welfare-related identity stigma decreased welfare enrollment levels by 5 percentage points. That is, persons who perceived levels of identity stigma one standard deviation below the mean had a 0.15 probability of enrolling in TANF, whereas those who perceived identity stigma at levels one standard deviation above the mean had a 0.05 probability of enrollment. The effect of welfare reform reduced the probability of TANF enrollment by 9 percentage points in states with more stringent welfare policies. Turning to Medicaid (Table 5), the results show that Medicaid-related identity stigma was not significantly associated with reduced Medicaid enrollment, but there was an important cross-program effect: Welfare-related identity stigma was associated with reduced Medicaid enrollment. Treatment stigma, enrollment barriers, and having less knowledge about Medicaid program rules also reduced Medicaid participation. In contrast to the effects of perceived poor treatment by program staff, concerns about provider treatment had no effect. Several policy variables

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

522 / Determinants of Participation in TANF and Medicaid

Table 3. Mean values of explanatory variables. Variable

TANF Medicaid Enrolled Not Enrolled Enrolled Not Enrolled

(N = 176) (N = 698) (N = 386) (N= 515) Identity stigma

Welfare stigma 2.99 3.27 3.07 3.29 Medicaid stigma 2.00 2.25

Treatment

Poor treatment while

stigma

applying and participating

in the program 1.55 1.77 0.76 1.14

Poor treatment by health

care

providers

2.25

Other individual determinants Enrollment

barriers

2.0

2.42

2.32

1.74

2.2

Confusion about eligibility 1.44 1.75 1.69 2.06

Misinformation about

Medicaid

rules

0.19

0.39

Need

Poor health 0.33 0.26 0.28 0.27 Number of children 2.35 2.04 2.16 2.04

County unemployment rate 4.46 3.90 4.12 3.91 Demographic variables Hispanic Black

Other

Age

0.22

0.55

race

0.03

31.2

Gender

0.37

0.26

0.04

32.0

0.23

0.41

0.45

0.20

0.04

30.0

0.04

33.1

0.96 0.89 0.93 0.88 0.63 0.62 0.62 0.62

Education

Number of parents 1.23 1.51 1.29 1.59

Annual income $3,723 $11,525 $6,301 $12,738 State policy Welfare

reform

Simplified

application

Generosity

1.20

1.69

Medicaid

process

of

the

1.38

3.06

1.77

2.51

Medicaid

program 1.02 0.81 0.93 0.80 Simulated eligibility 0.81 0.38 0.71 0.39

Gross income cutoff $10,432 $8,623 $13,533 $10,592 Interaction: income &

simulated eligibility $958 $947 $2,268 $2,050

increased enrollment, including simplification generous Medicaid benefits, and higher gross inc The mean probability of Medicaid enrollment w sample. Welfare-related identity stigma decrease centage points, while treatment-related stigma a both decreased Medicaid participation by 10 perc

edge about Medicaid program rules resulted in

Medicaid participation. States with more generou ment rates 25 percentage points higher than the enrollment policies increased enrollment by 13 p

This content downloaded from 129.186.138.35 on Mon, 10 Apr 2017 02:21:07 UTC All use subject to http://about.jstor.org/terms

Determinants of Participation in TANF and Medicaid / 523

Table 4. TANF enrollment models (N = 874).a b, Model 1

Model 2

Identity stigma Welfare stigma Treatment stigma

Model 3

Model 4

-0.37 (0.09)**

-0.27 (0.10)**

applying and participating

-0.11 (0.13)

-0.06 (0.09)

Confusion about eligibility

-0.11 (0.07) -0.08 (0.08)

-0.09 (0.07) -0.04 (0.07)

Poor treatment while

Other individual determinants Enrollment barriers Need

Poor health status 0.30 ( (0.23) Number of children 0.28 ( (0.08)** Unemployment rate0.66( (0.48)

0.52 0.36 (0.22)* (0.25)* 0.51 (0.21)** 0.31 0.26 (0.08)** (0.08)* 0.26 (0.08)** 0.03 (0.20) 0.83 (0.60) 0.003 (0.22)

Demographic variables Hispanic -0.12 (0.36)

Black 0.79 (0.19)** Other race -0.05 (0.42) High school graduate 0.21 (0.22)

Two parent household -0.31 (0.20)

Annual income -0.0001 (< 0.0001)** Age -0.003 (0.01) Female 0.96 (0.72) State policy

0.06 (0.34) -0.17 0.04 (0.30) (0.43) 0.67 0.71 (0.17)** (0.26)* 0.66 (0.16)** -0.22 -0.18 (0.33) (0.54) -0.21 (0.35) 0.18 0.18 (0.20) (0.23) 0.15 (0.19) -0.34 -0.24 (0.17) (0.27) -0.25 (0.19) -0.0001 (< 0.0001)** -0.0001 (< 0.0001) -0.0001 (< 0.0001) -0.0002 (0.01) -0.004 (0.01)* -0.002 (0.01) 1.13 (1.08)** 0.87 (0.49)* 0.85 (0.50)*

-0.38 (0.06)** -0.37 (0.07)** 0.57 (0.19)* 0.52 (0.21)** 2.12 (0.78)*" 2.17 (0.80)** 0.0001(< 0.0001)** 0.0001 (< 0.0001)**

Welfare reform

Medicaid generosity Simulated eligibility

Gross income cutoff

Interaction: income &

-0.0001 (< 0.0001)** -0.0001 (