| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
Social Security Administration, Office of Policy, Office of Research, Evaluation, and Statistics, Washington, DC.
Address correspondence to Kalman Rupp, Social Security Administration, Office of Research, Evaluation, and Statistics, 500 E Street, SW, 9th Floor, Washington, DC 20254. E-mail: kalman.rupp{at}ssa.gov
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionREFERENCES |
|---|
Methods. Using a microsimulation model and survey data matched to SSI administrative records, we model the effects of eight alternative policies on current and potential SSI recipients. We develop an evaluation methodology that systematically compares poverty outcomes, using multiple measures of effectiveness, at given levels of program expansion.
Results. All but two of the SSI reforms are clearly more target efficient at various degrees of simulated program expansion than popular proposals to reform Social Security. For a given cost increase, modifying the SSI asset test is the most effective option for reducing severe poverty among elderly women, but several reforms of the SSI income test are also highly effective.
Discussion. The SSI program is target efficient in providing a broad safety net to economically vulnerable elderly women. The relationship between SSI and Social Security and the relationship between the SSI asset and income tests have to be reevaluated to make the program more effective and appropriate to changing household structure and work patterns among the elderly population.
POVERTY has declined among the elderly population over the past 30 years, but the rate for elderly women remains almost twice as high as that for men. Factors contributing to lower Social Security benefits and higher poverty among elderly women include lower lifetime earnings, the breakdown of the nuclear family, fewer years spent in the labor force, relatively long life expectancy, lower likelihood of receiving pension income, and lower financial net worth. In addition, elderly women are less likely to be married than elderly men and more likely to be widowed or divorced. The death of a husband is followed by a decline in living standards and substantial reductions in wealth (Hurd & Wise, 1987
). Other developments may increase the risk of poverty among elderly women in the future. The ongoing secular shift from defined benefit plans to defined contribution plans means that retirement income is increasingly vulnerable to market risk and being cashed out upon job change. Social Security reform proposals involving individual accounts could add an additional element of market risk.
The progressive benefit formula embedded in the Social Security system involves a redistribution of income from men to women; however, a large part of these transfers are within families and thus do not generally accrue to poor women (Gustman & Steinmeier, 2000). Because of this, it is difficult for Social Security reform proposals to target benefits to the poor elderly population or poor elderly women. Anzick and Weaver (2001) look at five popular reform proposals focusing on the widow(er) benefit and enhancing Social Security's special minimum benefit. They estimated that only 14% to 37% of new government spending would go toward the reduction of poverty. Other proposals involve reallocating Social Security benefits from couples to surviving widows (Burkhauser & Smeeding, 1994; Iams & Sandell, 1998; Sandell & Iams, 1997), but they do not address the increasingly large group of never married elderly women. An unintended side effect of this type of reform is the loss of Supplemental Security Income (SSI) benefits (and Medicaid) for some widows (Kijakazi & Primus, 2000).
The SSI program, by contrast, can target benefits to poor elderly people generally and poor elderly women specifically quite efficiently. Conceived as a scaled-down negative income tax program focusing on the "deserving poor," it is a particularly important source of income for economically vulnerable women. Elderly people in poverty receive approximately 78% of all SSI benefits to the elderly population. Of these benefits, 75% go to elderly women (considerably more than half of all benefits paid to elderly people). This reflects the population share of poor elderly women relative to all poor elderly people (76%), as well as demographic and economic factors related to economic vulnerability among elderly women. Although SSI appears to be highly target efficient, only 34% of elderly women in poverty received SSI in 1998 (all figures are our calculations from the Survey of Income and Program Participation).
Can the SSI program be used to further reduce poverty among elderly women? The means-tested nature and the track record of the SSI program suggest that it can; however, the potential is limited by the structure of the program. The federal SSI income guarantee is at a subpoverty level—$552 for individuals and $829 for couples in 2003. State supplements bring the total above the poverty level in only a minority of states. In addition, the asset test cuts off SSI benefits (and often eligibility for Medicaid), even for individuals with no other countable income, at an arguably low level: $2,000 for individuals and $3,000 for couples. These cutoffs are not indexed to inflation, resulting in increasingly stringent criteria in real terms. Another limiting feature of the SSI program is that assets and income are treated differently, yet elderly people often trade assets for income through annuities.
Although the SSI program appears to be highly target efficient, it is unable to alleviate a large portion of poverty among elderly women. Both of these factors suggest excellent potential for SSI reform as a tool of further reducing poverty. However, it is not immediately obvious what kind of SSI reform is most target efficient and consistent with the demographic and economic trends of the 21st century. How should SSI handle Social Security reform proposals with unintended negative effects on SSI eligibility, such as proposals with a defined contribution component that could disqualify extremely poor people from SSI (Kijakazi & Primus, 2000)? Does it make sense to have an income guarantee that is relatively higher for couples than for elderly individuals living alone, while imposing a "marriage penalty" because cohabiting couples are treated as if they were individuals living alone?
We present and implement a new methodology that should be of help in considering the relative merits of alternative SSI reforms. The cornerstone of our methodology is a cost-equivalent comparison of alternative reform proposals and the use of a coherent set of measures to evaluate poverty outcomes. These features improve upon past studies in two ways. First, the cost of alternative reform proposals differs widely, making it impossible to say whether a costly intervention that reduces poverty substantially is more or less effective than an inexpensive alternative that reduces poverty less. (See the wide array of differences in cost and poverty reduction outcomes in Davies et al., 2002; McGarry, 2002; and Zedlewski & Meyer, 1989.) Second, the use of outcome measures is ad hoc and varies across studies. Given the inherent arbitrariness of aggregating individual welfare, it is more honest to provide a multidimensional view of poverty outcomes than to favor a single measure that is not inherently superior to other measures of distributional outcomes. We draw on a large literature on poverty measurement, particularly the seminal work of Sen (1976) and Foster, Greer, and Thorbecke (1984).
We present a set of SSI policy simulations involving reform proposals affecting both eligibility and benefits. Our results and discussion provide a useful step toward understanding the potential role of SSI in improving the safety net for elderly women while considering program cost and spending efficiency.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionREFERENCES |
|---|
Policy Interventions Tested
Modifications of the federal benefit rate (FBR)
The FBR is the amount of benefits (income guarantee) provided to an individual or couple with no countable income. The FBR is indexed to inflation and is separately established for individuals and couples. The FBR, in itself, does not provide poverty level income.
The first intervention, increasing the FBR, would affect all beneficiaries and would make some individuals and couples with low assets but incomes above the current FBR eligible for a small benefit (and Medicaid). Because it affects beneficiaries across the board, relatively small increases in the FBR may result in substantial increases in program cost.
Currently, the FBR for couples is 150% of the FBR for individuals. This is substantially higher than the official poverty line for a two-person family relative to a family of one. The second policy intervention would increase the FBR for individuals only. Because women tend to outlive their husbands and the proportion of widows is high among elderly women, this approach should allocate a large proportion of additional program funds to single elderly women.
Modifications of the general income exclusion (GIE)
The GIE applies to the first $20 of income. Any unearned income beyond $20 is offset by an equivalent benefit reduction. The GIE is not indexed to inflation; therefore, its value has eroded over time. The low value of the GIE is consistent with the social safety net nature of the SSI program. However, this results in the lack of recognition of past work efforts that may be reflected in Social Security payments.
The third intervention involves increasing the GIE. Without other changes, increasing the GIE would not affect the SSI benefits of elderly individuals with no countable income, who represent the poorest segment of SSI's target population. It would increase the benefits of current elderly SSI beneficiaries with income above the GIE limit, and it would make some people eligible who are currently ineligible because of high income.
As an alternative to the GIE intervention, a fourth intervention creates a Social Security income exclusion while leaving the GIE for sources other than Social Security unaffected. This intervention targets additional resources on those elderly people who have Social Security income, and it provides higher benefits to those who have higher Social Security incomes. The rationale is that past work effort reflected in Social Security benefits should be rewarded. In addition, the approach can be partly justified by the administrative simplicity of accounting for Social Security income.
Modifications of the earned income exclusion (EIE)
Under current SSI rules, any remainder of the GIE after consideration of unearned income can be used to exclude earned income. The EIE provides for the exclusion of an additional $65 of earned income, along with 50% of any remaining earned income. The fifth and sixth interventions increase the $65 earned income exclusion and reduce the 50% earned income tax rate, respectively. Because labor force participation among the elderly population tends to be low under current conditions, these interventions are expected to have only modest effects.
Modifications of the asset test
The asset rules are justified on the grounds that SSI provides a safety net of last resort; people with assets should spend them before turning to public assistance. Nevertheless, people with no or limited income may be ineligible if countable assets exceed the thresholds.
The SSI asset test has been criticized on several grounds. The current SSI asset test provides a strong incentive to spend down assets (Neumark & Powers, 1998) and eliminates the ability of recipients to preserve a fund for emergencies. In addition, it treats individuals differently depending on the form of their retirement savings. A person with a given level of income might qualify for SSI if the source of income is Social Security or an annuity, but this person might not qualify if the income source is an individual retirement account, a savings account, or an investment that can produce an equivalent income stream. As Kijakazi and Primus (2000) noted, this could become a major barrier to SSI participation as defined-contribution plans grow in popularity and under Social Security reform scenarios involving mandatory individual accounts.
The seventh intervention involves increasing the asset threshold. It preserves the current program approach in which the couple asset threshold is equal to 150% of the individual asset threshold. This intervention does not change the benefits or income of people who already qualify for SSI, but it allows people with low income and modest assets to qualify for SSI. Thus, this intervention might substantially increase the income of some of the poorest elderly.
The final intervention would transform the asset test into an integral part of the SSI income determination process, based on the notion that assets can be seen as equivalent to the income they are capable of producing. While formally eliminating the asset test, the intervention would create an income debit that counts x% of countable assets against the income test. The value of x can be theoretically justified or empirically determined. This intervention has similar properties as the previous one, except that it eliminates the discrete jump from benefits to no benefits associated with the asset threshold and allows people with much higher asset levels to qualify for SSI.
Data
The SSI eligibility and participation models are based on the 1990 panel of the Survey of Income and Program Participation (SIPP) matched to administrative data from the Social Security Administration (SSA). All estimates pertain to the 1991 noninstitutionalized U.S. population. The SIPP provides detailed data on the sources and amounts of monthly income and assets, which are crucial to the estimation of SSI eligibility. We also utilize detailed SIPP data on demographic characteristics and household composition. We replace survey measures of SSI participation and benefits with SSA administrative data, which are of high quality as a result of their central role in the administration of the SSI program (Huynh, Rupp, & Sears, 2002).
Eligibility and Participation Models
We focus on all elderly individuals who meet the SSI income and asset eligibility criteria. For individuals who live with their spouse, SSI eligibility also depends on the resources and income of the spouse. Consequently, we use the characteristics of "family units" rather than individual characteristics per se to estimate SSI financial eligibility.
We develop a probit model to estimate SSI participation for SSI-eligible elderly units. We derive the dependent variable, SSI participation status, from SSA administrative records. All other variables are taken from the SIPP. Following McGarry (1996), the principal determinant of SSI participation is the expected SSI benefit. Following Davies and colleagues (2002), we define expected SSI benefits as the sum of the potential federal SSI benefit calculated by the SSI eligibility model, plus the maximum state SSI supplement. Other covariates include demographic and economic characteristics of the unit head, such as age, gender, race or ethnicity, marital status, educational attainment, home ownership, place of birth (United States or elsewhere), U.S. citizenship, and health. We use average state Medicaid expenditures on SSI recipients to approximate the effect of Medicaid on SSI participation. A vector of state dummy variables is included to control for unmeasured state-specific effects on SSI participation.
Simulation Methodology
We estimate participation probabilities under the new program rules by applying the coefficients from the baseline participation model to the characteristics of eligibles under the new program rules. Base-case and simulation-case estimates of the number of participants are derived by summing the estimated participation probabilities across all eligible elderly women using sample weights.
The cornerstone of our methodology is a computationally feasible, iterative method of running a large number of simulations that converge upon a specified cost target. The parameter values from the simulations thus support comparisons of cost-equivalent policy alternatives. For example, if we are considering a 5% increase in program costs to be achieved by adjusting the FBR, the simulation iterates over potential values of the FBR until a value is found that is expected to generate a 5% increase in program cost. The same process and cost target is used for all other simulated changes to the SSI program parameters, thus producing a grid of cost-equivalent outcomes that can be compared against each other in a consistent fashion.
The simulations use a search algorithm to find the appropriate parameter values. The input is a predetermined level of program cost; the output is the value of the program parameter of interest, such as the FBR. The algorithm computes the simulated cost based on an initial guess of the parameter value and compares it with the predetermined target program cost. Then, the algorithm makes a linear adjustment to the parameter based on these two data points (the base case and the guess) and repeats the process. Specifically, the change in the parameter is computed to make the ratio of the desired change in the parameter (base case to target) to the actual change in the parameter (base case to guess) equal the ratio of the desired change in cost (base case to target cost level) to the actual change in cost (base case to guess). When the simulated cost is within $100,000 of the target cost, the algorithm accepts the parameter value and the iteration ceases. The $100,000 value represents 0.05% of baseline program costs.
The results we present focus specifically on poverty effects for elderly women; however, cost-equivalent policy changes are identified by considering program cost effects for all elderly individuals, including men. We have chosen this approach because we do not envision any of the policy alternatives to be applied to women only. We consider the robustness of our results with respect to gender, and we argue that final policy decisions should be made based on a consideration of the effects on both men and women. We also note that the poverty measures we report do not focus specifically on elderly women who participate in the SSI program. Rather, the base for the poverty measures is the sample of all elderly women. Although the SSI program has a large and increasing component targeting nonelderly disabled people (including children), in this article we focus on the elderly population alone. It would be worthwhile to address the robustness of our results with respect to the nonelderly population and to reassess whether uniform financial eligibility rules make sense for the elderly and disabled components of the SSI program. However, the disability component of the program raises complex measurement and analytic issues that merit a separate study. Our focus here is on the role of SSI reform in alleviating poverty among elderly women.
Our methodology makes a number of simplifying assumptions. Consistent with previous studies (Davies et al., 2002; McGarry, 2002), the model is static, does not incorporate potential behavioral effects of the interventions, and assumes that all changes occur instantaneously. The possibility of long-term behavioral changes would be most likely in the context of incentives for spending down assets or increasing work behavior, particularly in the 20% cost increase simulations. For the 1%, 3%, and 5% cost increase simulations, we do not expect behavioral effects to be of significant concern. Our estimates are conditional on the characteristics of the population underlying the simulations. The effects of the policy changes may be altered in the future if, for example, the size and composition of the target population change. In addition, we do not model the potentially complex interactions between the federal SSI program and state SSI supplements or other programs such as Food Stamps and Medicaid. The interventions examined by Zedlewski and Meyer (1989) generally result in trivial effects on Food Stamp expenditures. Medicaid is likely to be the most important element of non-SSI cost associated with the policy interventions we test. Finally, we note that the estimated outcomes are subject to uncertainty resulting from both sampling error and errors in predictions of participation.
Poverty Measures
Given the inherent arbitrariness of aggregating individual welfare, we provide a multidimensional view of poverty outcomes rather than focusing on a single measure that is not unambiguously superior to other available measures. We measure the poverty-reducing effects of each potential intervention based on the Foster, Greer, and Thorbecke (1984) class of poverty measures. The general formulation of the Foster, Greer, and Thorbecke (FGT) class of indices is
|
|
poverty rate (
= 0), normalized poverty gap ( = 1), and FGT index ( = 2).
One advantage of the FGT class of indices is that, although the choice of is somewhat arbitrary, the formulation of the indices is explicit. Moreover, the parameter represents poverty aversion. For = 2, greater weight is given to the poorest poor relative to those who are only slightly below the poverty threshold (Foster et al., 1984).
Each poverty measure described here has strengths and limitations. The poverty rate measures the proportion of a population with incomes below the official poverty threshold. The change in the poverty rate reflects only movements across the poverty line and is not sensitive to increases (or decreases) that occur in the income of the poor below the poverty line.
The poverty gap is the difference between the official U.S. poverty threshold and family income, summed across all elderly women below the poverty threshold. It can be interpreted as the minimum amount of money needed to eliminate poverty, assuming no behavioral changes. Although it captures the overall magnitude of poverty in a given population, the poverty gap ignores the severity of poverty by giving equal weight to a change in income for individuals just below the poverty line and individuals with the lowest incomes.
The FGT2 index is a poverty index that accounts for income inequality among the poor. It provides greater weight to poverty among the poorest poor, and the index value is comparatively low for interventions that aid those at the bottom of the income distribution. Although the choice of = 2 for the FGT2 index is arbitrary, it has the advantage of being mathematically related to the poverty rate and poverty gap. Any value of > 1 considers the severity of poverty. As increases, the weight given to the severity of poverty increases. In the limiting case as approaches infinity, the index collapses to a simple formula that is solely a function of the proportion of individuals with zero income.
When analyzing the results of the policy simulations, we present the percentage change between the base case and the simulation case in the poverty rate, the poverty gap, and the FGT2 index for elderly women. We also present figures on the proportion of benefits that would be received by poor elderly women, poor elderly men, and the nonpoor elderly population as an alternative way of demonstrating relative effectiveness.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionREFERENCES |
|---|
|
; McGarry, 1996, 2002; Zedlewski & Meyer, 1989). Over half of elderly female participants are widowed, over 84% have less than a high school education, and 34% are in poor health. Elderly female participants are more financially disadvantaged than elderly women as a whole. Although 74% receive Social Security income, the median monthly benefit among recipients is only $278. The average SSI benefit among elderly female participants is $181 per month (not shown). Table 1 also shows that eligible female participants are slightly less likely to be in poverty than all eligible women, reflecting the success of the SSI program in reducing poverty. The last three columns of Table 1 present data for elderly women who are not eligible for SSI benefits. The columns for elderly women who are income eligible only and resource eligible only show the characteristics of those who might become eligible for SSI as a result of the interventions we test. Those who meet only the SSI income eligibility criteria have low median SSI-countable monthly incomes, low median Social Security income, and high median SSI-countable assets. Their poverty rate (84.2%) is the highest of all groups, suggesting that these elderly women would benefit greatly from interventions targeting the SSI asset test. Elderly women who meet only the resource eligibility criteria have low median SSI-countable assets, high median SSI-countable monthly incomes, and high median Social Security income. Although this group has a relatively low poverty rate, it represents over 3.5 million elderly women who might benefit from interventions targeting the SSI FBR or income exclusions. By contrast, those elderly women who meet neither the SSI income nor resource eligibility criteria have a poverty rate of only 2.1% and have very high median SSI-countable income and assets. Thus, the SSI eligibility criteria are extremely effective in screening out elderly women who are clearly not poor.
Table 2 presents the results of the participation model, with bootstrapped standard errors to account for the complex SIPP sample design. The key result is that expected SSI benefits are positively and significantly related to the likelihood of SSI participation. The marginal effect indicates that a $100 increase in benefits would increase the overall probability of participation by 5 percentage points. The U.S. born appear to be more likely to participate than immigrants. U.S. citizenship, however, is negatively related to SSI participation. Thus, because everyone born in the United States is a citizen, the combined effect is small at best. Among immigrants, the participation probability varies by entry cohort. The data suggest that obtaining U.S. citizenship offsets the possibly higher SSI participation among immigrants. Eligibles who require personal care and who had a greater number of doctor visits in the previous 12 months are more likely to participate. The estimated coefficients on the state dummy variables are not presented because of space constraints.
|
|
|
The simulated parameter values for the Social Security income exclusion are similar to the simulated values of the GIE at each cost level. In fact, the Social Security income exclusion acts as a targeted increase in the GIE. The inflation-adjusted value of the simulated Social Security income exclusion, even for a 20% cost increase, is much lower than the exclusion proposed by some. For example, Smeeding, Estes, and Glasse (1999) proposed a $175 Social Security income exclusion. We estimate that a $60 Social Security income exclusion (in 1991 dollars) would increase SSI program costs for the elderly population by 20%. This is roughly equivalent to the $80 exclusion proposed by Kijakazi and Primus (2000) after inflation is accounted for.
Table 5 presents the implications of the simulations in terms of poverty among elderly women. It presents the percentage change between the base case and the simulations for the poverty rate, poverty gap, and FGT2 index. The results are grouped by outcome measure. Each column represents a specific change in federal SSI program costs.
|
The second panel of Table 5 illustrates the changes in the poverty gap that correspond to each intervention. In contrast to the poverty rate, the changes in the poverty gap vary across interventions and the magnitudes are substantial. The interventions are not all equally effective, however. The two simulations involving the asset test stand out from the rest in terms of reducing the poverty gap. This is true at the 1%, 3%, and 5% levels of cost increases. It is not feasible to change the asset test to produce a 20% cost increase; a simulation of the complete removal of the asset test produced a cost increase of only 8% (not shown). The large effects associated with changing the asset test are remarkable in light of the results of previous studies. For example, eliminating the asset test has the smallest poverty gap reduction of any of the changes simulated by McGarry (2002). We find that, on a cost-equivalent basis, the simulations involving changing the asset test display comparatively impressive results.
A general pattern of the relative effectiveness of the various interventions in reducing the poverty gap can be discerned from Table 5. Behind the two involving the asset test is a second group of interventions that includes increasing the FBR, increasing the individual FBR only, increasing the GIE, and creating a Social Security income exclusion. Finally, a third group of interventions includes the two simulations involving the EIE, which are clearly the least effective. These patterns hold at all feasible cost levels.
The third panel of Table 5 shows percentage changes in the FGT2 index. A negative change indicates a decrease in poverty. The interventions can be grouped into the same three categories as for the poverty gap. That is, the two interventions involving the asset test are the most effective, there is a middle group of somewhat less effective interventions, and the two interventions involving the EIE are the least effective. Thus, using a measure that emphasizes income changes at the bottom of the income distribution does not change the ordering of the results; however, the asset test interventions stand out even more when the FGT2 measure is used.
Although the results discussed herein are for elderly women only, we argue that final policy decisions should be based on a consideration of the effects on both men and women. Table 6 presents the proportion of additional dollars that would be received by poor elderly women, poor elderly men, and nonpoor elderly people (for the 5% cost increase scenario). Consistent with previous results, the two interventions involving the asset test would direct a relatively large proportion of additional spending to elderly women and men in poverty. The two earned income exclusion interventions would largely direct additional spending to elderly people who are not in poverty. The ranking of the overall target efficiency of the various interventions (proportion going to poor women and men combined) is the same as the ranking for women alone. Thus our results are highly robust by gender. Table 6 also shows that the proportion of additional spending going to poor elderly women and men for each of the interventions is largely a function of the gender distribution of the elderly poor. The one exception to this is the intervention that increases the federal benefit for individuals only, which disproportionately targets poor elderly women. This reflects the high proportion of widowed women among the elderly poor as well as the growing proportion of never-married elderly women.
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionREFERENCES |
|---|
What is the potential of SSI reform for enhancing the social safety net for elderly women in the 21st century? To answer this question it is necessary to put SSI reform in the broader context of Social Security reform and other important changes to society and the economy.
Perhaps the Achilles' heel of successful Social Security reform is the need to find cost-effective solutions to counter economic vulnerability among the elderly population that may arise from personal accounts allowing for investments in the stock market, the shift from defined benefit pension plans to defined contribution plans (also partially invested in stocks), and the secular shift to a society of unrelated or loosely related individuals. By contrast, our study has found that SSI reform can be structured in a highly target-efficient way. Social Security reform in this respect is inherently limited for several reasons. It does not address the economic vulnerability of population subgroups outside of the Social Security system. Moreover, it does not address poverty arising from poor investment outcomes from individual accounts. Reforms focusing on establishing a minimum Social Security benefit are inherently limited because the structure of Social Security is closely tied to earnings histories—those with sporadic work histories may not be sufficiently protected under such proposals. Proposals focusing on widows are limited to a subgroup, albeit an important one, but also one that is becoming less important as a result of the breakdown of the nuclear family and the increase in female employment. Moreover, reforms related to Social Security alone are poorly targeted (Anzick & Weaver, 2001) because they ignore other sources of income and assets.
The major attraction of SSI for the elderly population as a tool for fighting poverty is that it is a negative income tax targeted directly on poor elderly people. Thus looking at Social Security and SSI reform together is sensible, all the more so given that the financing of the two programs is fundamentally different—trust fund versus general revenues. Viewed in isolation, some reforms may appear costless because they simply shift costs to a different pot of funds.
The current SSI program, however, is not as effective at fighting poverty as it could be. One of the most important reasons is the asset test. From an antipoverty standpoint, the current SSI asset test has two major weaknesses. First, rather than gradually phasing out SSI benefits as asset levels increase, there is a cliff at the threshold, a threshold that is very low relative to the SSI income guarantee. Second, there is no consideration given to the income-producing potential of financial assets other than passbook savings and dividend-producing assets. Thus eligibility for SSI benefits depends not only on the value but also the form of financial assets. These problems may not have been substantial in the 1970s, a time when few people invested in the stock market and defined benefit plans were dominant. Both of these issues will become increasingly important in the decades ahead, especially if Social Security reforms oriented toward individual equity investments are implemented.
Given these problems, it is no surprise that our empirical results show that poverty among elderly women is by far the highest among those who are income eligible, but have assets above the asset threshold. We tested two asset reform options, both of which are highly target efficient. The first is to simply increase the asset threshold. This approach, however, does not eliminate the threshold cliff or address the income-producing potential of assets. The second is to "tax" assets by converting asset values to an imputed income stream that is counted against the FBR, while eliminating the threshold requirement—the cliff. On a cost-equivalent basis, the second approach tends to be slightly better in reducing poverty among elderly women. It also has appealing incentive properties. In addition, a reasonable argument can be made to set the asset "tax" rate close to the annuitized value of the asset.
Another important reform approach involves changing the economies of scale assumptions embedded in the SSI benefit formula. This is addressed by the reform to increase the individual FBR only. As expected, this approach consistently produced better results for elderly women than increasing the FBR across the board. The couple FBR is 150% of the individual FBR. If the individual in question is living alone, this implies a FBR for couples that assumes weaker scale economies than are embedded in the official poverty threshold measure.
There is also an SSI marriage penalty arising from the fact that the program does not currently acknowledge economies of scale arising from two or more unmarried adults living in the same household. Thus, whereas the couple FBR is 150% of the individual FBR, a cohabiting couple, if both qualify, would receive benefits calculated on the basis of 200% of the individual FBR. A promising direction for research might involve simulations designed with an eye toward the better alignment of the FBR applicable to persons in different household structures and the economies of scale assumptions of poverty measures.
Finally, we address the trade-off between targeting and economic incentives, an issue that is particularly important in the context of savings and work behavior. One of the attractive aspects of asset test reform is that it would simultaneously improve targeting and reduce perverse incentives to "spend down" assets. However, as with any means-tested income support program, there are clear trade-offs between targeting and economic incentives. Most strikingly, although our results show that liberalizing the EIE is an ineffective tool of reducing poverty among the elderly population, it would undoubtedly increase work incentives. Similarly, although creating a separate SSI exclusion for Social Security income may be slightly less target efficient than increasing the GIE, a Social Security income exclusion should reduce lifetime work disincentives. The work incentive issues may become increasingly important as life expectancy increases, the Social Security normal retirement age is raised, and increased work activity among elderly people is facilitated by technological change. Thus, although target efficiency should be a prime objective of SSI reform, other factors such as work and savings incentives, as well as administrative complexity, also have to be considered.
|
Acknowledgments |
|---|
The authors jointly prepared the article. The order of the authors is random and does not indicate relative importance. The views and opinions expressed in this article do not represent the official position of the U.S. Social Security Administration or any other entity.
|
Footnotes |
|---|
Received for publication September 2, 2002. Accepted for publication March 25, 2003.
|
REFERENCES |
|---|
|
TOPAbstractMethodsResultsDiscussionREFERENCES |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
