The Availability of Benefits

Disability benefits are not provided on the basis of trust but through complex disability determination processes that administer the conditions required to receive benefits (Stone, Reference Stone1986). These conditions include, most prominently, the verification of whether a person’s impairment meets the threshold for a work-disability as defined by a particular disability benefit program. The compensation index evaluates this dimension by comparing cross-national policies pertaining to, for example, the minimum required work-incapacity level to receive benefits, the strictness of the medical and vocational assessment criteria, and the maximum disability benefit payment level provided (Böheim and Leoni, Reference Böheim and Leoni2017). A survey-based approach, on the other hand, seeks to analyze data on the outcomes of these policies directly from the population being studied. Survey data representative of the geographical area being studied are, of course, a prerequisite to this approach. With a representative survey in hand, the availability of benefits can be explored by identifying both the population of those with work-disabilities and the population receiving disability benefits. Benefit availability can then be assessed based on the share of the population with work-disabilities receiving disability benefits. A comparative reference group can further be used to evaluate the degree to which benefits are available relative to other contexts.

Identifying the population with work-disabilities requires consideration as to how disability is defined. While historical definitions of disability focused on the presence of a medical impairment (Marin, Reference Marin and Prinz2003), modern notions of disability, including the social model of disability and the International Classification of Functioning, Disability and Health (ICF), view disability as the gap between a person’s functional limitations and their unaccommodating environmental circumstances (Brandt et al., Reference Brandt, Houtenville, Huynh, Chan and Rasch2011; Shakespeare and Watson, Reference Shakespeare and Watson1997; WHO, 2001). As Mont and Loeb (Reference Mont and Loeb2010: 163) explain: “Knowing only that a person has had polio tells you very little about their degree of disability, the quality of their life, or their ability to participate in productive activities.” These researchers thus developed a functional limitations indicator, combining questions on activity limitations and participation restrictions, that is consistent with ICF notions of disability. Croda and colleagues (2018) compared the performance of a number of health indicators at predicting disability benefit enrollment and found that the functionality index of Mont and Loeb (Reference Mont and Loeb2010) was the strongest predictor of disability benefit receipt cross-nationally. The index is formed following a principal components analysis based on multiple functional items, including indicators for mobility, fine motor, and health limitations, and difficulties completing the activities and instrumental activities of daily living.

The strengths of the functionality index are that it uses multiple data points to identify those with functional limitations and is theoretically grounded in the ICF. The limitation is that it requires large amounts of data on functional limitations which may not always be collected, particularly in surveys focusing on younger adults. Information on functional limitations may also not directly indicate the inability to engage in the specific participatory domain of work. To address this issue, prior research has identified those with work-disabilities as those who scored in the bottom quintile of the functional index and who were also not in work, not retired, and who self-reported a work-limiting disability (Morris, Reference Morris2021). A concern with this approach to identifying work-disability is that the self-report of work-disability can be endogenous to the labor market and welfare state context (O’Brien, 2013; Reference O’Brien2015). The self-report of work-disability may further be biased if respondents exaggerate the severity of health problems to rationalize their decision to apply for disability benefits (Benítez-Silva et al., Reference Benítez-Silva, Buchinsky, Man Chan, Cheidvasser and Rust2004). Surveys may also assume that those with work-disabilities necessarily receive disability benefits. For example, in the U.S. Health and Retirement Study (HRS), which provides a nationally representative sample of adults ages 51 and older in the U.S., until 2016 only those who reported a health-related work limitation were asked questions on SSD benefits (Hyde and Harrati, Reference Hyde and Harrati2021). With these self-reported work-disability issues, analysts may simply not include this variable while using other labor force variables, such as whether the respondent reports being in work or retired, in conjunction with the functionality index to identify those with presumed “work-disabilities” among the working-age population.

In addition to a robust indicator of work-disability, data on the receipt of disability benefits are critical for the analysis. Self-reported information on disability benefit receipt is widely collected in social surveys, which provides an easy option for researchers though it also presents limitations. For example, the HRS survey questions on disability benefit receipt, including on whether the respondent receives SSI or DI, are not consistently asked across waves. It is also important that the survey questions reflect updated terminologies and program characteristics that are adopted to ensure accuracy, particularly when programs have undergone reforms. An additional limitation with the use of self-reported benefit receipt data is that respondents may confuse the different social welfare programs and provide faulty information. For example, both the old-age insurance and disability benefit programs are commonly referred to as “Social Security” in the U.S. and thus some respondents may mix up these programs (Hyde and Stapleton, Reference Hyde and Stapleton2017). Respondents could also confuse the receipt of private disability insurance with the receipt of public disability insurance.

For these reasons, researchers must be cautious in the analysis of self-reported beneficiary data. Increasingly, survey data linked to administrative records are available to researchers via restricted access channels (Svedberg et al., Reference Svedberg, Ropponen, Lichtenstein and Alexanderson2010; Hyde and Harrati, Reference Hyde and Harrati2021). This represents an important pathway to obtaining verifiable information on disability benefit receipt as the data come directly from the governmental source administrating the program. Linked administrative data, however, will only include survey participants who have consented to the release of their administrative data. This can bias the analysis as it will not include data on those who did not consent. Thus, while linked administrative data hold much promise for identifying reliable information on who receives disability benefits, the issue of missing data is an important one that must be addressed.

The need for a reference point to evaluate the availability of benefits is another important methodological consideration. As it is unreasonable to expect that all people with work-disabilities will receive disability benefits (or obtain full decommodification), a comparative reference group is required to evaluate performance. That reference group can take multiple forms. It might include the evaluation of cross-sectional data on the share of those with work-disabilities who receive disability benefits and how that has changed over time or over a specific policy reform period. Banks and colleagues (2015), for example, examine changes in the probability of receipt of disability benefits for those in poor health before and after the introduction of a new disability benefit program in Britain. The reference group could also be a different country or welfare state regime (see Morris, Reference Morris2021; Croda et al., Reference Croda, Skinner, Yasaitis and Lindeboom2018). Having a comparative reference point that is either historical, cross-nationally comparative, or both will enable researchers to evaluate the extent of disability-related decommodification as it pertains to the availability of benefits.

The Generosity of Benefits

The second core pillar requiring exploration in our decommodification research framework is the generosity of benefits. To realize disability-related decommodification requires not only that benefits are available but that they provide enough benefits to provide an adequate standard of living. In the decommodification index (Esping-Anderson, Reference Esping-Anderson1990), this is measured by calculating a replacement rate based on a percentage of the average production worker’s wage covered by benefits. In the compensation index covering disability benefit programs, it is similarly measured through the maximum disability benefit payment level as a replacement rate of 75% or higher (Böheim and Leoni, Reference Böheim and Leoni2017). These measurements notably base the decommodification standard on the replacement rate for the average worker without disabilities. People with disabilities, however, often incur higher costs of living due to the need to purchase disability-related goods and services. For instance, across 15 Organisation for Economic Cooperation and Development (OECD) countries it is estimated that workers with disabilities require, on average, 44 percent more income to make ends meet relative to comparable adults without disabilities (Morris and Zaidi, Reference Morris and Zaidi2020). Though one could fruitfully adjust the compensation index for the extra costs of disability, the analysis of survey data provides a more direct means of evaluating the generosity of disability benefit programs and whether these programs enable beneficiaries to sustain a reasonable living standard. As with the analysis of benefit availability, such an analysis requires a reliable indicator of benefit receipt. It also requires a suitable indicator of living standards and consideration as to how a reasonable living standard is analyzed.

Quantifying the amount of benefits received from disability benefits and then comparing that to the income received by workers without disabilities or to the income of the beneficiaries themselves prior to their disability onset represent compelling options. However, an approach focusing on self-reported income-replacements has its limitations. Income provides a poor proxy for living standards, particularly as it relates to people with disabilities who encounter extra costs of living (Zaidi and Burchardt, Reference Zaidi and Burchardt2005). Moreover, the quality of the data on self-reported benefits and income can be error prone as it relies upon respondents to recall this information (Alwin et al., Reference Alwin, Zeiser and Gensimore2014). An alternative approach, widely used in the extra costs of disability literature, is to measure living standards through a composite indicator of assets or subjective measures of deprivation or financial insecurity (Morris and Zaidi, Reference Morris and Zaidi2020). In utilizing these deprivation measures, the analyst is able to produce a living standard indicator that is sensitive to the conversion disadvantage experienced by people with disabilities and the problem of treating income as a direct measure of utility. The challenge, however, is that the choice of a standard of living indicator is subjective, may vary across people with different preferences or tastes, and can also differ in terms of the availability of questions across surveys. An index of the standard of living that combines different questions can help to create a more robust indicator that “averages out” differences in preferences or tastes (Zaidi and Burchardt, Reference Zaidi and Burchardt2005). It is also important that the standard of living indicator be valued equally by people with and without disabilities. The ability to afford health care, for example, is likely to be of greater value to people with disabilities than people without disabilities and would thus make for a poor standard of living indicator, whereas questions relating to being able to pay monthly bills or to put food on the table will be better as they likely represent a living standard of equal value to those with and without disabilities (Zaidi and Burchardt, Reference Zaidi and Burchardt2005).

Once a reliable indicator of living standards is established, researchers can proceed to analyze the extent to which disability benefit recipients are able to uphold the standard. Analysts can calculate the share of disability benefit recipients who are unable to obtain the established living standard relative to those not receiving disability benefits. The analysis can be further refined using multiple regression to identify whether disability benefit recipients experience greater difficulty meeting the standard of living compared to adults who are not on benefits but with similar socio-demographic characteristics. As with the analysis of benefit availability, comparative analysis can also be conducted to inform an analysis of the extent of achieved decommodification. Such a comparative analysis, however, requires that the living standards indicator be harmonized cross-nationally, which is information that is generally less likely to be standardized than health related information.Footnote ¹

Policy Background

In the U.S., the two major disability benefit programs, DI and SSI, share a definition of disability “as the inability to do any substantial gainful activity by reason of any medically determinable physical or mental impairment which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months” (SSA, 2021b). The DI program, enacted in 1956, provides income replacements to insured adults medically determined to meet the disability definition and with a record of working and paying Social Security taxes. In 2019, DI workers, on average, received $1,384 per month in cash benefits (SSA, 2021a). The SSI program, enacted in 1972, pays cash benefits to adults and children with disabilities on the basis of financial need in order “to meet basic needs for food, clothing, and shelter” (SSA, 2021c). Unlike DI, to receive SSI benefits applicants must not surpass a resource limit ($2,000 for an individual or $3,000 for a couple) and must also have low income (maximum $841 a month for an individual). In 2019, SSI eligible individuals received a monthly maximum of $794 in cash benefits (SSA, 2022).Footnote ² DI and SSI beneficiaries are also eligible for health insurance: Medicare for DI beneficiaries after receiving benefits for 24 months and Medicaid for SSI recipients.Footnote ³ Both programs are not time-limited – however, beneficiaries may be subject to periodic continuing disability reviews.

Data

To evaluate the degree to which these programs support disability-related decommodification, we draw on the HRS, which provides a nationally representative longitudinal sample of adults ages 51+ in the United States and which collects a wealth of information on health, demographics, and economics of aging. While the analysis below is thus restricted to older beneficiaries, the majority of DI beneficiaries and SSI disabled adults are older than 50 (SSA, 2021a). HRS respondents are surveyed every two years about their employment, health, income, and a range of other topics. We restrict the analysis to HRS respondents ages 51 to the full retirement age in the U.S. so as to capture adults eligible for SSD benefits of working-age. We further draw on various special features of the HRS available to researchers, including the user-friendly RAND HRS file providing cleaned and consistent variables across waves, access to restricted administrative data on benefit receipt obtained following ethics review approval and a formal application process, as well as the ability to compare the HRS results with the ex-ante harmonized Survey of Health, Ageing, and Retirement in Europe (SHARE). The SHARE, which provides a data set covering the health and well-being of older adults across 27 countries, is modeled after the HRS with about two-thirds of the variables in SHARE identical to those available in the HRS (see Börsch-Supan et al., Reference Börsch-Supan, Bucher-Koenen and Hanemann2020). As in prior work (see Börsch-Supan et al., Reference Börsch-Supan, Bucher-Koenen and Hanemann2020; Morris, Reference Morris2020; Croda et al., Reference Croda, Skinner, Yasaitis and Lindeboom2018), the SHARE and HRS datasets are used to cross-nationally compare the dynamics of disability benefit receipt.

In examining the availability of benefits, we analyze a restricted HRS data file that includes linkages at the individual level to sources of administrative information on SSD benefit receipt. Specifically, we use the Disability Analysis File which contains historical data on HRS respondents since 1996 who participated in the DI and SSI programs and who consented to allow the HRS to obtain their administrative benefits records. Second, we compare results pertaining to the availability of long-term disability benefits in 27 other high-income countries using the SHARE. With uniform questions in both surveys, identical measures of work-disability are constructed in the HRS and SHARE allowing us to compare the proportion of the older-adult working-age population with work-disabilities that receives disability benefits across 28 countries, including the U.S.

To evaluate the generosity of benefits, we draw on the so-called Leave Behind Questionnaire (LBQ) of HRS, which was first introduced in the 2006 wave. The LBQ, which does not have a cross-nationally harmonized question base in SHARE, is administered to half of the HRS respondents each wave, which occurs every two years, and thus during every two-wave period every respondent in the HRS is surveyed. Using respondent level identifiers, the LBQ file can be merged into the RAND HRS file. For the purposes of this study, we draw on the 2016 wave of the LBQ which will be useful as it contains questions concerning financial insecurity that serve as a proxy for the standard of living.

Evaluating the Availability of SSD Benefits in the U.S.

This section examines the extent to which the U.S. SSD programs effectively target benefits to those with work-disabilities. We draw on longitudinal data from the HRS linked to administrative data from the SSA. Given the variation in the entry cohorts into the HRS of each wave and to ensure a full sample of older adults ages 51 to the entitlement age, we examine four HRS samples for the years 1998 (birth years 1936 – 1947), 2004 (birth years 1938 – 1953), 2010 (birth years 1944 – 1959), and 2016 (birth years of 1950 and younger). We constructed the latent index of functional ability of Mont and Loeb (Reference Mont and Loeb2010) across the four time periods using 20 functional related questions. The list of the variables and the principal component scores constructed in the index are provided in Table A1 in the online appendix. A work-disability indicator is constructed consisting of respondents in the bottom functional health quintile, who were not retired, and who were not in work.

Though self-reported SSD information is included in the publicly available HRS modules, the administrative data allow us to identify benefit recipients with greater precision. The major limitation of the administrative data, however, concerns the lack of consenters, which leads to missing data preventing a nationally representative sample. To account for this limitation, we use information on the consent status of all HRS respondents and follow the approach of Hyde and Harrati (Reference Hyde and Harrati2021) by weighting for those who provide consent which we apply in the results below. We generated the weight by running a logistic regression in each sample year using numerous sociodemographic variables to predict the likelihood of consenting. These predicted values are then used to generate inverse probability weights which are applied with the RAND HRS sampling weights to the sample of consenters.

Table 1 presents the weighted descriptive statistics indicating the share of those who receive SSD benefits, the share with work-disabilities, and the share of those with work-disabilities who receive SSD benefits across four time periods. The information is further presented using both the administrative data on disability benefit receipt and the self-reported disability benefit receipt data. A first finding concerns the general similarities between the administrative and self-reported data. This suggests the general reliability of the self-reported benefit information. We do observe, however, a slightly higher share of those with work-disabilities receiving SSD benefits in the administrative data compared to the self-reported benefit information. A second finding concerns the growing share of the population with work-disabilities receiving disability benefits over time – from 32% in 1998 to 47% in 2016 using the administrative data. It is notable that this increase is consistent with a growing share of beneficiaries as indicated in the table. Prior research has closely scrutinized the reasons underlying this trend, which include population ageing, increasing DI eligibility for women, declining mortality, and an increasing number of allowances for mental and musculoskeletal conditions (Liebman, Reference Liebman2015). These results suggest that this growth in benefit enrollment corresponded with an increase in the receipt of disability benefits among those with work-disabilities. A third finding is that more than 50% of those with work-disabilities, people who presumably stand to benefit from these programs, do not receive disability benefits.

Table 1.

Comparing administrative and self-reported data on disability benefit receipt

Notes: Self-reported data reported using RAND HRS weights. Administrative data reported using author-constructed weights described in the text. We provide unweighted figures in Table A3 and conduct statistical tests to determine significant differences in the probability of those with work-disabilities receiving disability benefits in Tables A4 and A5 in the appendix. p<.05 = *; p<.10= + indicates statistically significant change relative to 1998.

In the recent recalculation of the compensation index covering disability benefit policies cross-nationally, the U.S. was found to have one of the least generous compensation policies among 20 OECD countries (Böheim and Leoni, Reference Böheim and Leoni2017). We thus hypothesize that a lower portion of people with work-disabilities in the U.S. would receive benefits compared to those in other OECD countries. In Figure 1, we test this hypothesis comparing the estimates of the availability of benefits to a cross-nationally harmonized sample of working-age older adults across 27 additional high-income countries. The SHARE study was designed to foster cross-national comparability with the HRS. Identical work-disability indicators are thus constructed for 27 countries in the SHARE among comparable samples of older adults below state pension ages in the specific countries (OECD, 2017). The self-reported receipt of the main or secondary disability insurance pension in the given country further serves to indicate the receipt of disability benefits in the specific country. These comparative results are presented in Figure 1, which include the U.S. estimate from the HRS using the self-report of disability benefits for comparative purposes. Here we see that the percent of the working-age older adult population with work-disabilities receiving disability benefits at 44% in the U.S. is slightly above the average across all 28 countries, which stands at 41%. This suggests, counter to our hypothesis, somewhat greater availability of disability benefits to those with work-disabilities in the U.S. relative to other countries.

Figure 1.

Share of working-age older adults with work-disabilities receiving disability benefits.

Note: HRS data for the U.S. covering the year 2016 and SHARE wave 7 data for the other countries covering the year 2017. All values weighted using survey and country-specific weights. Full principal component results available upon request.

Evaluating the Generosity of SSD Benefits

We next evaluate the generosity of SSD benefits applying the approach described in the research framework using the LBQ sub-sample of the HRS restricted to those age 65 and younger. To do so, we develop a composite measure of financial security which serves as a proxy for the standard of living. Specifically, three questions on financial security were used and re-coded so that a higher score indicated greater financial security. These questions included: how difficult is it for you/your family to meet monthly payments on you/your family’s bills (coded as 1 for complete difficulty to 5 for not difficult at all); how satisfied are you with your present financial situation (coded as 1 for not at all satisfied to 5 for completely satisfied); and the report of ongoing financial strain (coded as 1 if reports no ongoing financial strain and 0 if reports strain). These questions are consistent with the research framework as they represent universally valued objectives that can be formed into a composite measure. We create the measure by summing the three items into a financial security index (ranging from 2 for low financial security to 11 for high financial security). The scale demonstrated good scale reliability (Chronbach’s alpha of 0.82). The self-reported benefit receipt information is used for this analysis as the sample is limited to half the HRS respondents who respond to the LBQ thus leading to greater missing data in the administrative information. While a useful measure for illustrating the proposed framework, the measure is limited in a number of respects. It is possible, for example, that some benefit recipients report obtaining financial security but do so only by not purchasing needed disability-related goods and services. The financial security index, however, cannot identify such unmet needs. Moreover, the index provides no insight into the kinds of deprivations experienced due to financial reasons, such as inadequate housing or nutrition. The measure is also self-reported and thus may be viewed as subjective.

Prior research matched information from SSA administrative records to analyze the income levels of DI and SSI recipients in 2013 and found that nearly 20 percent of DI recipients and 43 percent of SSI recipients had family income below the federal income-based poverty threshold (Stegman and Hemmeter, Reference Stegman and Hemmeter2015). Based on this research, we hypothesize that both DI and SSI beneficiaries will experience a lower likelihood of achieving financial security relative to non-beneficiaries with similar socio-demographic characteristics, thus indicating the limited generosity of benefits. Given the lower benefit amounts and strict asset requirements of the SSI program, we further suspect that SSI beneficiaries will experience lower amounts of financial security compared to DI recipients. We therefore conduct the subsequent analysis for DI and SSI recipients separately. The full summary data on the sample evaluated are reported in Table A2 in the appendix and indicates higher average financial security index scores for non-SSD beneficiaries (7.66) relative to those who receive SSD benefits (5.94). We also observe lower mean financial security scores for SSI recipients (5.05) compared to DI recipients (6.02). Table 2 Models 1 and 2 report the bivariate relationships between DI and SSI receipt on financial security and indicate that, compared to non-beneficiaries, DI recipients score, on average, 1.33 points lower (β = −1.325, p = .000) on the financial security index, while SSI recipients score 1.68 points lower (β = −1.680, p = .000). When introducing controls in Models 3 and 4, the likelihood of lower scores on the financial security index for DI (β = −0.990, p = .000) and SSI (β = −1.079, p = .002) recipients remains statistically significant.

Table 2.

Regression of financial security on DI and SSI benefit receipt

Note: Analysis of HRS Leave Behind Questionnaire 2016. Controls include work-disability, years of education, age, number of children, gender, Hispanic, White, Black, census region. Full results available upon request. As a sensitivity test, we also test these relationships using a Tobit model that adjusts for the censuring of the dependent variable. These results did not substantially change the findings and are available upon request. * p<0.05, ** p<0.01, ***p<0.001.