Indicators of Welfare Dependence, 1997

Chapter I.
Introduction

The Welfare Indicators Act of 1994 (Pub. L. 103-432) directed the Secretary of Health and Human Services to develop indicators of the extent to which American families depend upon income from welfare programs. Welfare programs, as defined under the Act, include the Aid to Families with Dependent Children (AFDC) program, the Supplemental Security Income (SSI) program, and the Food Stamp Program (FSP). Under the Welfare Indicators Act, annual reports are to be made concerning:

• The rate at which families depend on income from these welfare programs;
• The degree and duration of welfare recipiency and dependence;
• Predictors of welfare dependence; and
• Additional data needed to assess issues relating to welfare dependence.

An Advisory Board on Welfare Indicators also was established under the Act to assist the Secretary in defining welfare dependence and in choosing appropriate data for inclusion in the annual reports. The Board consisted of a bipartisan group of experts appointed by the Senate, the House of Representatives and the President. The Board oversaw the production of the Interim Report to Congress, published last October, and they played a major role in shaping this report, which is the first of the annual reports to Congress required under the law.

This report addresses the requirements of the Welfare Indicators Act in several ways. In this first chapter, the specific summary measures of welfare dependence proposed by the Advisory Board are presented and discussed. These measures attempt to provide a small set of indicators that can be tracked routinely to monitor progress in reducing welfare dependence. It is anticipated that they will be published on an annual basis in much the same way that poverty measures, for example, are published. At this point the measures are still somewhat experimental and further comments and discussion are invited.

The second chapter of this report presents a broader group of indicators of welfare recipiency and dependence. These indicators include measures of the extent of recipiency for each of the three programs considered separately, as well as information on income from all three programs in combination. Interaction of AFDC, SSI and FSP benefits with periods of employment and with benefits from other programs are also shown. The chapter also includes data on movements onto and off welfare programs, and on the extent to which welfare recipiency in adolescence is correlated with later adult recipiency.

The Board expressed a strong view that dependence measures could not be assessed in isolation, since changes in these measures could result either from increases in work activity and other factors that would raise family incomes, or from sanctions in welfare programs that would reduce welfare program participation but might not improve the material circumstances of these families. Accordingly, they recommended that measures of deprivation such as poverty rates, with and without the inclusion of welfare benefits, be presented together with the dependence measures. This chapter follows that recommendation and presents data on several measures of deprivation over time periods corresponding to those shown for the recommended measures of dependence.

Chapter III focuses on "predictors" of welfare dependence -- risk factors believed to be associated with welfare receipt in some way. These predictors are shown in three different groups: those that concern families' degree of economic security, those that are related to the work status of adult family members, and those that relate to teen behaviors.⁽¹⁾ Economic security -- including measures of poverty, receipt of child support, health care coverage, and so forth -- is important in predicting dependence in the sense that families with fewer economic resources are more likely to rely on welfare programs for their support. Factors related to work status are also important, because families must generally receive an adequate income from employment in order to avoid dependence without severe deprivation. And finally, teen behaviors are very important since a high proportion of long-term welfare recipients became parents as teens, often outside of marriage. Starting a family in these circumstances may lead to dependence because teens generally lack adequate skills, preparation and resources to support a child.

Chapter IV addresses the final goal set out above by discussing additional data that might be needed to construct better indicators and predictors. Although the measures included in this report are the best that could be constructed with currently available data, additional data would potentially provide greater insights into the problems associated with welfare dependence.

Further, data needs are likely to change over time as welfare programs change. Since the passage of the Welfare Indicators Act, significant changes have been made in the federal system of providing means-tested assistance to families. The Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) (Pub. L. 104-193), enacted in 1996, increased state flexibility and gave states considerably more freedom to operate their own assistance programs. These changes mean that variations across welfare programs are likely to increase greatly, and our measures may need to be revised substantially to reflect the many different things that states may do to assist needy families. Chapter IV discusses this problem in greater detail.

Welfare recipiency is a necessary pre-condition for welfare dependence, although the Advisory Board cautioned that the two are not the same thing. As discussed above, welfare programs have changed substantially in the recent past, and are continuing to change rapidly. Appendix A has therefore been included to give basic data on each of the three main welfare programs and their recipients over the past several years.

Measuring Welfare Dependence

Measuring welfare dependence is not a simple problem. How much income must be received from welfare programs before a family is considered "dependent?" Does dependence relate to the length of time that the family has received benefits, or only to the amount or proportion of income from welfare? Are families with some income from work dependent if they also receive welfare income? These are among the questions that the Advisory Board considered in attempting to define and measure welfare dependence.

This report retains the concept underlying the Interim Report -- that dependence is a continuum, with variations in degree and in duration. Families could be more or less dependent if larger or smaller shares of their total resources were derived from welfare programs. Further, the Advisory Board recommended that a family receiving one welfare check probably should not be classified as dependent, even if that check accounted for all of their income for the month in question. Dependence, in other words, has some inherent concept of recipiency in more than the very short term. Finally, the Board believed that income associated with work, even if it was ultimately provided by the public, should not be counted as assistance in estimating dependence.

Although many different measures of dependence could be constructed that would follow these guidelines, the Board recognized the need for a summary measure that could be tracked over time to give some broad indication of changes in the overall degree of welfare dependence in American society. They developed a particular definition of welfare dependence that fulfilled that need. For this purpose, they proposed that the following definition be put forward for discussion and debate:

A family is dependent on welfare if more than 50 percent of its total income in a one-year period comes from AFDC/TANF, Food Stamps and/or SSI, and this welfare income is not associated with work activities. Welfare dependence is the proportion of all families who are dependent on welfare.

This measure is not without its limitations. The Advisory Board recognized that no single measure could fully capture all aspects of dependence and that their proposed measure should be examined in concert with other key indicators of dependence and deprivation. In addition, while the proposed definition would count unsubsidized and subsidized employment and work required to obtain benefits as work activities, currently available data do not permit distinguishing between welfare income associated with work activities and non-work-related welfare benefits. As a result, the data shown in this report overstate the incidence of dependence (as defined above) because work required to obtain benefits is classified as welfare and not income from work. Neither does this proposed definition capture all aspects of dependence. In particular it represents an essentially arbitrary choice of a percentage (50 percent) of income from welfare beyond which families will be considered dependent. However, it is relatively easy to measure and to track over time, and is likely to be associated with any very large changes in total dependence, however defined. For example, as the recent changes in welfare law move more recipients into employment or work-related activities, dependence under this definition should be expected to decline.

Unfortunately, any such declines cannot yet be observed using currently available data. The most accurate data on sources and amounts of income across the population as a whole are from the Survey of Income and Program Participation (SIPP).⁽²⁾ Because the SIPP is a large longitudinal survey of households it takes longer to collect and process than do program case records, and therefore representative data on all families appear somewhat later than do case record data.

Table SUM 1 shows the percentages of families who receive any welfare benefits and the percentage who would be considered welfare dependent under the above definition for the most recent years for which data are available.⁽³⁾ There is little trend discernable in these data. While there have been small year to year changes in both recipiency and dependence, the changes seen in the data available so far are not large enough to be statistically significant even in a survey as large as the SIPP.⁽⁴⁾ Overall, between four and five percent of all individuals would be considered welfare-dependent based on these data. These families represent about one-third of those who receive any benefits in each year.

Table SUM 1. Percent of the Total Population with More than 50 Percent of Income from Means-Tested Assistance Programs

Table SUM 2. Percent of the Total Population with More than 50 Percent of Income from Various Means-Tested Assistance Programs, 1992

The Advisory Board's discussion also focused on the need for some measure of dependence that included the concept of recipiency over an extended period of time. Accordingly, they recommended the inclusion of an additional measure that considered what proportion of the population participating in welfare programs over various periods of time met the dependence criteria. Table SUM 3 summarizes that measure for two different time periods, 1972-1981 and 1982-1991.

Table SUM 3. AFDC Receipt and Percentage of Recipients with More than 50 Percent of Income from AFDC and Food Stamps by Number of Years

Note: "AFDC Receipt" is defined as whether the person received AFDC at any time during the year. "AFDC & Food Stamps >50% of Income" is defined as whether the person's AFDC and Food Stamps benefit was more than 50% of their yearly income. "0 Years" means that while the person received means-tested assistance, their benefits were 50% of their income for zero years during the time period. For example, a person listed as receiving AFDC for 6 - 8 years ("Any AFDC Receipt") may never have received benefits greater than 50% of their income (0 years, AFDC and Food Stamps >50% of Income).

Source: Unpublished data from the PSID, 1972 - 1991.

Even among families who were recipients, the majority were dependent on AFDC and Food Stamps for less than one year in total over each of these 10 year periods. For example, 11 percent of people who received welfare at all received it for 9 to 10 years, but only 4 percent of those who were dependent on welfare at any point were dependent for 9 to 10 years. As the spell of recipiency lengthened, it appears that recipients were more likely to supplement welfare with income from other sources such as earnings. There is a small tendency for the proportion of spells of welfare dependence that are longer to grow over this period, but the change is not large enough to be statistically significant.⁽⁵⁾

Measuring Deprivation

Measures of dependence may change for a number of different reasons, both positive and negative. As discussed earlier, the Advisory Board cautioned that measures of dependence should be presented in context -- that is, with some measure of the impacts of dependence changes on deprivation. Many different measures could again be used for that purpose.

One measure of deprivation is to look at changes in the level of need over time. Elsewhere in this document, for example, measures of the "poverty gap" (see Appendix B) -- the amount of income that would be needed to bring all of those below poverty to the poverty line -- and of food insecurity are presented (see Chapter III). Both of these give some indication of changes in the level of need over time. Further, both focus on changes that affect the resources of the part of the population that is already classified as poor. This is appropriate in considering effects of changes in welfare programs, because most welfare recipients have below-poverty incomes even including their cash welfare benefits.

In this chapter, however, the deprivation measure presented focuses directly on changes in the anti-poverty effectiveness of welfare and related programs. Tables SUM 4 and SUM 5 (and their associated figures) show how much welfare programs have reduced poverty rates over the period since 1979, first for all persons and second for persons in families with related children under age 18.

These tables show that many more families would be poor if they did not receive welfare benefits. Counting only cash income (excluding welfare), the poverty rate would generally be four to five percentage points higher than it is calculated to be when means-tested cash benefits, food and housing benefits, and taxes (including refunds through the Earned Income Tax Credit (EITC)) are all counted. This final poverty rate -- taking into account all sources of support -- is a more complete measure of deprivation than is the official poverty rate or other measures that exclude some types of support. Breaking it down in this fashion allows the relative contribution of different sources -- including cash welfare and relatively fungible in-kind welfare benefits -- to the alleviation of poverty to be observed.

Poverty rates of all types began to increase in 1990 as the economy went into a recession, reaching a peak in 1993. As economic conditions have started to improve rates have come down, both before and after means-tested assistance. Poverty rates for families with children remain substantially higher than those for all families, however. The gap between poverty rates before and after public assistance has increased slightly over time, particularly in the last few years as the size of the EITC has grown. The EITC is a work-related benefit, however, and is not included as assistance in estimating dependence. Through 1995 the contribution of means-tested welfare programs to the reduction in poverty has remained roughly constant at about four percentage points, although during the recession of the early 1980s these programs did somewhat less to reduce total poverty. Current poverty-reduction rates for assistance programs are about the same as in 1979, although a bit more of the reduction comes in the form of non-cash benefits.

The relatively small changes in the level of overall deprivation since the late 1980s is consistent with the small changes in the dependence rate seen earlier. As larger changes in dependence occur under PRWORA, it will be both necessary and interesting to track changes in these deprivation rates as well. If this legislation succeeds in its aims, dependence should fall noticeably while deprivation measures remain largely unchanged.

Table SUM 4. Trends in Poverty with and without Means-Tested Benefits for All Persons, 1979 - 1995

Note: The first measure of poverty, labeled cash income plus all social insurance, includes social security but not means-tested cash transfers. Adding means-tested cash transfers yields the official census definition of poverty, the second line in the table. Food and housing benefits may be received either as cash or (more generally) as in-kind benefits in which case the market value of food and housing benefits is added. EITC refers to the refundable Earned Income Tax Credit which is always positive whereas Federal payroll and income taxes are a negative adjustment. The fungible value of Medicare and Medicaid is not included.

Source: Congressional Budget Office tabulations. Additional calculations by DHHS.

Table SUM 5. Trends in Poverty with and without Means-Tested Benefits for All Persons in Families with Related Children Under Age 18, 1979 - 1995

Note: The first measure of poverty, labeled cash income plus all social insurance, includes social security but not means-tested cash transfers. Adding means-tested cash transfers yields the official census definition of poverty, the second line in the table. Food and housing benefits may be received either as cash or (more generally) as in-kind benefits in which case the market value of food and housing benefits is added. EITC refers to the refundable Earned Income Tax Credit which is always positive whereas Federal payroll and income taxes are a negative adjustment. The fungible value of Medicare and Medicaid is not included.

Source: Congressional Budget Office tabulations. Additional calculations by DHHS.

1. Because there are so many potential predictors and not all possibilities can be included here, additional details on potential risk factors and associated indicators are provided in Appendix B.

2. While the number of families dependent on welfare from a particular program could potentially be calculated using program data alone, the proportion of the total population who are welfare dependent could not be seen, because families may receive income from more than one program and it is not always possible to match records across programs to avoid double-counting such families. Additionally, studies have shown that reports of total income to household surveys such as the SIPP are typically more detailed and accurate than are reports to welfare program administrators. Finally, the definition of a "welfare unit" for program purposes does not always include all family members living in the same household.

3. While more recent data from the SIPP have been collected, due to a number of technical issues, they were not available for analysis at the time this report was drafted.

4. Standard errors can be calculated using the formula published in the Survey of Income and Program Participation Users' Guide.

5. For further discussion of standard errors for PSID estimates, see The Panel Study of Income Dynamics, A User's Guide.

Where to?

Top of Page
Contents of Report

Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning & Evaluation (ASPE)
U.S. Department of Health & Human Services (HHS)

Updated: 11/29/01