Promoting research on Social Security policy

[michiganretirementresearch]

Building a better model of how retirement, health, and consumption interact

MRRC researchers John Karl Scholz and Ananth Seshadri recently built a model to look into how health and consumption intertwine, using an Euler equation to examine consumption in relation to utility. In this instance, consumption represents spending and health investments, such as eating well or exercising, and utility relates to health. In their model, “a retiree maximizes lifetime utility by making decisions on consumption and health investments in each period. The utility flow in each period depends on both consumption and health capital. Health investments could affect both the probability of surviving to the next period and future health capital conditional on survival. There are health shocks in each period such that both survival and future health capital conditional on survival are [randomly determined].”

The researchers applied their model to Health and Retirement Study (HRS) and Consumption and Activities Mail Survey (CAMS) data. Incorporating health into the model is the project’s contribution, but the authors’ findings also include:

Consumption generally slows as people age, but health shocks play an important role in slowing down decline

Including health investments in the model could help explain the wide differences in consumption-age profiles related to health.

Health investments, such as physical exercise, have a significant effect on the evolutions of both health and consumption in retirement.

The authors state: “The findings in this paper suggest that public programs like Medicare and Medicaid play an important role in shaping the consumption and health behavior of the retirees, as well as the retirement and saving decisions of working people.” Quantifying the magnitude of these effects might inform the optimal design of such programs.

Read the research brief.

Read the working paper

[michiganretirementresearch]

International Labour Organization posts two job opportunities

The ILO has two positions posted: One for a technical specialist in the Social Security Development Branch of the International Social Security Association. The incumbent would organize surveys of international social security organizations, research social protection schemes, write country profiles, and coordinate the maintenance and updating of databases.

ILO also seeks an economist for its research department to investigate the effects of labor market policies in terms of employment and social outcomes. Notably how the design of these labor market and social policies contribute to improving labor market outcomes, conditions of work and the well-being of populations around the world.

Both limited-term positions are based in Geneva, Switzerland, and have application deadlines of September 18, 2016.

For more info/to apply for the technical specialist job

For more info/to apply for the economist job

[michiganretirementresearch]

2016 RRC Meeting slides and papers are now available online

Thank you to all who presented at and attended this year’s Retirement Research Consortium meeting. If you missed it, or would like to refer back to slides and/or summaries, they are now posted on the MRRC website. Links below.

2016 RRC Book (with summaries)

Panel 1 slides: Health, Health Insurance, and Choice of When to Retire

Panel 2 slides: Cognitive Health

Panel 3 slides: New Ways to Insure Adequate Resources for Retirees

Panel 4 slides: Government Finances with an Aging Population

Panel 5 slides: Cohort Changes

Panel 6 slides: Household Resources in Old Age

Friday lunchtime speaker, Axel Börsch-Supan slides: “What’s up in Europe?”

Panel 7 slides: International Comparisons

Look for extended coverage of the meeting in the next MRRC newsletter.

[michiganretirementresearch]

Nearing retirement, higher home equity distinguishes U.S. immigrants from natives

With the wave of immigrants admitted post-1965 Immigration and Nationality Act nearing retirement age, David Love and Lucie Schmidt (LS) use their 2015 working paper, ”Comprehensive Wealth of Immigrants and Natives,” to ascertain whether immigrants in the United States are well prepared for retirement, relative to native-born residents of similar socioeconomic status.

Existing literature indicates that immigrants are less likely to receive Social Security benefits, in spite of making contributions, since they often fail to meet the eligibility requirements of having worked and paid Social Security taxes for a cumulative 10 years. They also have lower private pension coverage because they are less likely to work for firms that provide pension plans. However, comparing similarly situated immigrants and natives reveals that the former have significantly higher private net worth, and larger levels of home equity due in part to residential location choices that concentrate immigrants in areas of high housing cost such as New York and San Francisco. Therefore, while immigrants may not have sufficient pension and benefits coverage post-retirement, they may be able to compensate with private wealth.

LS assess retirement resources for households with respondents age 51 years and older using data from the Health and Retirement Study panel for the years 1998 to 2012. They also use restricted data on Social Security earnings and geography—country of residence and state of residence.  

In order to assess their level of retirement preparation, LS build comprehensive wealth measures for households, incorporating the value of a household’s financial and nonfinancial assets, as well as their annuitized benefits from Social Security and defined-benefit pensions. This provides a more accurate picture of household wealth than simple net-worth measures. However, there is the additional question of how this wealth will be spread over the time period following retirement. LS assume that this wealth is used to purchase an annuity that delivers a fixed income till end of life—annualized wealth.

Comparing wealth so constructed, LS find that while natives consistently hold more comprehensive wealth over all age brackets, immigrants draw down their wealth more slowly and, therefore, close the gap between themselves and natives at higher ages. Annualized wealth is similarly higher for natives, but the gap between natives and immigrants closes with age because of faster growth of housing wealth among immigrants.

LS investigate the source of the disparity between native and immigrant annualized wealth by exploring demographic characteristics that may be tied to these differences. They find that being in a dual immigrant marriage is associated with lower annualized wealth. However, despite exploring a very rich set of demographic characteristics, including earning data and country of origin, they are unable to pin down precisely what factors lead to the disparity between native and immigrant wealth. They also find strong cohort effects in the data, with earlier immigrants showing greater wealth accumulation in the data relative to immigrants in the recent, post-1975 waves.

LS also find that, consistent with earlier literature, the annualized wealth of households is increasing over time because of the growth in home equity. LS create simulated trajectories of annualized wealth to ascertain what assumptions are required to replicate the upward wealth trajectory that they find in the data. They discover that this trajectory is driven by the desire to leave bequests at end of life, and the desire to avoid the transaction cost of selling a house, which results in individuals retaining their home equity into very late ages.

The authors conclude that while immigrants appear to be well-prepared for retirement overall, more recent waves of immigrants appear less prepared with lower levels of annualized wealth than their predecessors. Policymakers may want to take note of this trend as it suggests greater economic vulnerability among immigrant populations going forward.  

Read the research brief.

Read the working paper.

Contact MRRC.

[michiganretirementresearch]

Live stream the 2016 Retirement Research Consortium meetings

For those who are unable to attend the RRC meetings in person, live streaming of the talks will be available.

The webcast for Thursday, August 4, will run from 8 a.m.-4:30 p.m. Eastern time. Thurday’s link: https://scontent.webcaster4.com/web/RRC1

The webcast for Friday, August 5, will run from 8 a.m.-2:30 p.m. Friday’s link:https://scontent.webcaster4.com/web/RRC2

The schedule is posted here. If you can make it to the Washington, D.C., Press Club, there’s still time to register here.

[michiganretirementresearch]

Aging brains want to know: What causes irrational financial choices?

Economic research has documented a negative relationship between older age and proficiency in financial decision-making. In their earlier work, Syngjoo Choi, Shachar Kariv, Weiland Muller, and Dan Silverman [CKMS] show that, in a choice experiment, adults older than 55 are more likely to make departures from economic rationality than their younger counterparts. By further showing a large positive correlation between economic rationality and wealth, the authors are able to make a strong case that such cognitive decline can have significant economic consequences.

In order to design appropriate policy to address the decline of decision-making abilities in old age, we must establish the mechanism by which this decline occurs. In their 2015 working paper, “Sources of Lower Financial Decision-making Ability at Older Ages,” Shachar Kariv and Dan Silverman [KS] set out to do this by replicating and extending the experiment in CKMS on a larger sample for whom detailed information on health and cognitive function is available.

The specific questions they want to address are:

Given that the experiments used to establish these results are web-based, does the decline in rationality only arise for older cohorts that are potentially more uncomfortable with the use of Internet and computer technology? If this is the case, the decline will not persist for cohorts who have been raised in the era of computer technology, and can be corrected in the older cohorts by making them more familiar with new technologies.

Is this decline in rationality distinct from other sources of cognitive decline seen with age? If so, policy can be tailored to directly address this decline in financial decision-making abilities, which may not be captured by standard policies to address aging and cognitive decline.

Is this decline driven by illnesses that are more likely to affect the elderly, or does it occur for all the elderly regardless of health? If driven by overall health shortfalls, we need to engineer policy toward those most health deficient or susceptible to such illnesses.

To explore these questions, the authors employ the Longitudinal Internet Studies for Social Sciences (LISS) panel, retaining all 4,160 panel members for whom demographic information is available. The participants are presented with 25 problems designed to measure whether their choices satisfy the Generalized Axiom of Revealed Preferences (GARP). This test of rationality assesses whether participants make consistent choices that satisfy utility maximization under different prices and budgets. They also calculate the Critical Cost Efficiency Index (CCEI) that quantifies the degree by which choices deviate from GARP, and the Money Pump Index (MPI) which estimates the degree to which these violations of consistency can be exploited in the market place.

After accounting for the effects of income, employment and other basic demographics, KS find that participants older than 64 have lower CCEI and MPI scores, indicating lower rationality relative to lower age participants. They find that older cohorts do appear to have a greater tendency to make errors that relate to their facility with the computer interface. However, this correlates very minimally with their lower rationality in the experiment, and therefore isn’t the underlying explanation for the errors in consistency shown by the data.

KS also employ Raven’s Matrix tasks to establish overall cognitive decline. If the cognitive decline measured through the Raven’s tasks is strongly correlated with GARP violations, this would suggest that the decline in economic rationality is really just a feature of overall decline in cognitive abilities and not a unique outcome in itself. They find, however, that this is not that case and lack of proficiency in Raven’s tasks does not explain the decrease in economic rationality. They further find that general measures of self-reported health and disability status don’t correlate with the decline with rationality either.

KS conclude, therefore, that the decline in economic decision-making does not have its roots in poor overall health, natural diminution in cognitive abilities with age, or cohort effects arising from lack of experience with computer technology. Policymakers may want to consider this decline of financial decision-making in the elderly as a concern in its own right to be addressed in designing financial products relevant to this population.

Read the paper.

Read the research brief.

Contact MRRC.

[michiganretirementresearch]

How do military benefits impact veterans’ retirement decisions?

Upon completion of 20 years of service, military retirees become eligible for a lifetime annuity  equal to half their basic pay, and increasing to 75 percent of their pay for 30 years of service. In addition, they have access to the military health-care system, Tricare. Consequently, the 2012 fiscal obligations toward nondisabled military retirees amounted to more than $42 billion in retirement benefits, excluding health-care costs. This has precipitated discussions on the need for reform of the military benefits system. In their 2015 working paper, “The retirement and social security benefit claiming of U.S. military retirees,” David Knapp, Beth Asch, Jim Hosek, and Michael G. Mattock (KAHM) suggest that there may be labor-supply implications for policy changes that impact these benefits.

KAHM model retirement benefits’ impact of on retiree labor supply as having an income effect by which higher, nonwage income from military benefits post-retirement creates incentives to retire early from subsequent careers. On the other hand, military benefits might allow retirees to engage in longer searches by providing them with supplemental income, resulting in a better job match and longer work duration because of higher pay and/or greater nonpecuniary benefits such as job satisfaction. Having supplemental benefits from the military might conversely result in retirees investing less in their job search because they already have a stable source of non-wage, retiree income. Therefore, military benefits can theoretically contribute to increased or decreased job duration in post-military careers.

In order to identify the impact of military benefits on labor-force participation, KAHM exploit a 2001 congressional act that extended Tricare benefits to Medicare-eligible military retirees. Without this reform, retirees would have had to purchase Medicare supplemental insurance, Medigap, at a 2001 cost of $1,524 per person. The extension of Tricare reduced deductible expenditures for people older than 65, as well as expected deductible expenses in the future for those younger than 65.

KAHM use data from the 1992 cohort of the Health and Retirement Study (HRS). The median age of this cohort would be 65 in 2001, when Tricare was extended. Their subsample for analysis consists of all individuals who were working at age 55. For this subsample, KAHM estimate how the probability of retirement changed as a result of being eligible for Tricare extension. Their comparison group is military veterans who are ineligible for Tricare extension and, therefore, should be unaffected.

Their results indicate that, if you are a military retiree reaching the age of 65 after 2001, you are more likely to retire after Tricare extension, relative to military veterans who are ineligible for Tricare. However, this difference is not statistically significant so we cannot reliably claim that the two groups are experiencing different trends in retirement post 2001.

KAHM explain, however, that by including all ages in their analysis, they are mixing the effect on people who are not yet 65 with those older than 65. The former expect a decrease in their medical expenditures in the future, but the latter actually experience this decrease and are more directly affected. On a restricted sample of 65 to 69 year olds, they find that the extension of Tricare increases the likelihood of retirement by 15.2 percentage points, a statistically significantly difference from veterans who are ineligible for Tricare. Their analysis would imply that military retirees are highly sensitive to the level of military benefits, with greater benefits resulting in earlier retirement.

KAHM extend their analysis to assess how military benefits affect Social Security benefits claiming. They find that Tricare extension results in a higher likelihood of military retirees delaying claiming, but this relationship also is not statistically significant so a robust claim regarding the effects of military benefits on claiming decisions cannot be made.  

KAHM caution that their results apply only to a small subset of military retirees, since retirees with 20 years of service may be as young as 38 if they enlist at 18 years of age. They suggest possible extensions of this research using administrative data on military retirees to include younger populations not captured by the HRS data.

Read the research brief

Read the working paper

Email MRRC

[michiganretirementresearch]

Mama may have, Papa may have: Dynastic income has little effect one’s own consumption

The marginal propensity to consume (MPC) is a metric of great importance to macroeconomic policy since it determines the amount of consumption spending that will result from policies that increase income. Theory and empirical work suggest that liquidity-constrained households – households that have limited ability to borrow and are therefore constrained in their ability to spend – have a higher MPC relative to other consumers. But what other features of households can be expected to affect their propensity to consume?

HwaJung Choi, Kathleen McGarry, and Robert F. Schoeni (CMS), in their 2015 working paper, “Liquidity Constraints the Extended Family, and Consumption,” address the question of whether resources of extended family members also have an effect on the MPC of liquidity constrained households. Hence, is MPC highest when both households and their extended family are liquidity constrained?

The inspiration for their work is derived from an influential paper by Joseph G. Altonji, Fumio Hayashi, and Laurence J. Kotlikoff (AHK). The AHK paper shows that the extended family’s income can affect a household’s consumption, even when the extended family constitutes a distinct household. However, AHK’s analysis was limited to food consumption owing to data limitations in the Panel Study of Income Dynamics (PSID) during the time period of their study. CMS are able to extend this analysis to broader categories of consumption because of an expansion in consumption questions asked in later years of the PSID. CMS use data from 1999-2011, which has questions covering 70 percent of household expenditures, and data from 2005-2011, which allows them to capture 95 percent of expenditures.  

The altruistic model of consumption suggests that household consumption is determined not by own income, but by the average resources of the extended family, referred to as dynastic income. The life-cycle model, on the other hand, holds household income to be an important determinant of household consumption. CMS find that their data predicts the same results as AHK, with a strong and statistically significant relationship between consumption and own income. Moreover, own income has a stronger relationship with nonfood consumption. Dynastic income is found to have a significant impact on consumption of food and other expanded categories of expenditures. The data is, therefore, inconsistent with a pure altruism model which would argue that only dynastic income affects own consumption and own income has no additional effect.

Having established a relationship between own income and consumption and dynastic income and consumption, CMS proceed to evaluate the MPCs for different wealth categories. As predicted by theory, the bottom quartiles and people in the 25 percent to 75 percent wealth percentiles have larger MPCs, and own wealth has no significant effect on consumption in the top wealth quartile. This pattern holds for food and nonfood consumption.

Also consistent with the researchers’ expectations, the bottom wealth quartile has the largest MPC when dynastic wealth is also low. However, this result is not replicated in the higher wealth categories. Therefore, if own wealth is in the 25th to 75th percentiles, or in the upper quartile, having low dynastic wealth does not result in a higher MPC.

CMS assert that it is important to understand the role of extended family in supporting individuals and households, particularly with the recent economic downturn and the increasing pressure on the Social Security program in providing for an aging population. Their data provides suggestive evidence for extended family wealth’s role in influencing household consumption, and for this effect extending to transitory income changes through the MPC.  

Read the working paper.

Read the research brief.

Contact MRRC.

[michiganretirementresearch]

Which job characteristics might keep us working?

Existing research has shown that there are correlations between job characteristics and the decision to retire. Physical and mental stress, as well as repetitive working conditions, increase the likelihood of early retirement, while workers in jobs requiring complex tasks and less physical exertion appear to delay retirement. However, data limitations have kept this body of work from addressing factors such as pension arrangements, individual health, cognitive traits and, most significantly, the relative importance of objective job characteristics and subjective perceptions of working conditions on retirement decisions. The literature addressing transitions to part-time work, as opposed to full retirement, is also limited.

In their 2015 working paper “Nonmonetary Job Characteristics and Employment Transitions at Older Ages,” Marco Angrisani, Arie Kapteyn, and Erik Meijer [AKM] expand on the literature by using richer data to provide a more comprehensive picture of the job attributes that affect labor-force transitions from full-time to part-time work and retirement.

DATA

AKM combine data on older adults from the Health and Retirement Study (HRS) with information on job characteristics provided by the Occupation Information Network (O*NET). The HRS data provides information on basic demographics, labor force participation, pension arrangements, health status, and financial situation. In addition, it collects information from respondents about job attributes such as physical and mental requirements, stress, incentives and pressures to retire, etc. This information is used as the basis for subjective job characteristics. The HRS also collects information on personality traits in every second wave, which AKM use to construct indices corresponding to the Big Five personality traits. The HRS also elicits planned retirement age and perceived probabilities of working full time after age 62 and 65 from its respondents. AKM investigate these as outcomes that might be affected by job characteristics.

The O*NET data is used to obtain “objective” job characteristics. By matching O*NET occupations to U.S. Census occupation codes that are employed in HRS job classifications, AKM are able to merge O*NET job characteristic to HRS occupations. O*NET characteristics are very disaggregated, therefore AKM develop indices of related characteristics under the broad themes of 1) qualifications and interests of the typical worker; 2) worker requirements; and 3) job activities and demands.

FINDINGS

AKM use the data to estimate the effect of nonmonetary job characteristics on the probability of employment transitions, the planned age of retirement, and subjective probability of working at older ages.

AKM find that job transitions from full-time to part-time work and retirement are strongly associated with nonmonetary job characteristics. However, objective characteristics of jobs collected through O*NET are more strongly related to the probability of transitioning into retirement, whereas subjective, self-reported job characteristics are better at predicting transitions between full-time and part-time work. Among objective measures, working with computers and working in an environment with greater social interaction reduce the probability of retiring, while greater physical exertion and diminished eyesight increase retirement probability. Similarly, higher subjective measures of physical exertion relate with increased probabilities of transitioning from full- to part-time work, while use of computers predicts lower probability of transitioning to part-time work. Perceived age discrimination at work is also strongly linked to both the probability of transitioning to part-time work and retirement.

Physical effort, and all variables that correlate with it such as eyesight and use of equipment, predict a lower planned age of retirement, and lower probability of working full time. Work requirements such as social skills, and high levels of concentration on the job prolong time till retirement and probability of working at older ages. Task difficulty and job stress increase probability of earlier retirement, but also increase the probability of staying in full-time work past 65. Perceptions of age discrimination are linked to earlier retirement and lower probability of full-time work at older ages.

Overall, AKM find that job characteristics developed from O*NET data are stronger predictors of actual retirement and retirement intentions, with subjective measures performing favorably in the prediction of labor transitions from full-time to part-time work. They caution, however, that given the nature of estimation involved, it is difficult to establish causal relationships from job characteristics to outcomes. There are unobservable characteristics of individuals that may also be related to these outcomes, and may be driving the effects seen in their data. However, this is a useful exploratory step toward establishing how nonmonetary characteristics of jobs can relate to people’s retirement choices.

Read the working paper.

Read the research brief.

Contact MRRC.

[michiganretirementresearch]

Examining the impacts of RET reforms on old-age poverty

By withholding Social Security benefits for earnings above an income threshold, the Retirement Earnings Test (RET) impacts the labor supply decisions of individuals who are earning near to and above this threshold. The RET also impacts the age at which beneficiaries start claiming benefits by creating an incentive to delay claims and, consequently, increase monthly receipts. Reforms to the RET in 2000, which led to the elimination of RET for people past Full Retirement Age (FRA) and made it less stringent for people in the year they achieve FRA, were intended to improve employment and earnings among people who may be deterred from working owing to reduced benefits. These reforms also created incentives to start claiming benefits early and, therefore, receive lower monthly benefits overall. In their 2015 working paper “Does Eliminating the Earnings Test Increase the Incidence of Low Income among Older Women?,” Theodore Figinski and David Neumark (FN) investigate the net impact of these reforms on the income of older women, ages 75 and older. FN assess whether removing RET may in fact be increasing old-age poverty.

Prior research has shown that eliminating the RET threshold results in increases in labor supply among men and women, with earnings increases of 19 to 20 percent. On the other hand, claiming benefits one year early has been shown to create a benefit reduction of 6.7 percent, with the reduction persisting for every successive year. This may quickly overwhelm the effect of short-run earnings increases. FN select older women as the appropriate sample for researching these opposing effects since women have higher life expectancy than men, and are, therefore, more likely to reach the age where they are dependent solely on Social Security benefits. This makes women more likely to be negatively affected by their or their spouse’s early claiming.

DATA

In an ideal world, we could estimate the effect of RET reform by comparing women’s outcomes after the reforms with their outcomes if these reforms had never occurred, but this counterfactual does not exist in actual data. FN instead compare different cohorts of older women with one another, on the basis that some cohorts were affected by the reforms and others were not. Moreover, there are differences in how long each cohort was exposed to the reforms, with younger cohorts being exposed for longer. Their primary samples consist of women ages 70 to 71 and 75 to 76 in the Health and Retirement Study (HRS) panel dataset. They also match women to husbands to explore the effect of spouses’ claiming decisions on the women’s outcomes in later age.

RESULTS & ANALYSIS

Their results show that the elimination of RET reduces the age at which women start claiming benefits by eight to nine months, with larger effects for younger women who were exposed for longer to the reform. Hence, women who were 65 at the time of the reforms claim 8.2 to 9.6 months earlier relative to estimates of 6.5 to 7.8 months for women who were 69. They find similar effects for spouses in the data, with men also lowering their age of claiming in response to the RET elimination. Their analysis proceeds to show that this reduction in the age of claiming results in a reduction in annual benefit claims, in both the individual sample and the family sample (own and spousal benefits).

FN proceed to evaluate the impact of the RET reforms on poverty incidence by estimating the likelihood of women living below 150 percent or 200 percent of the poverty line. They find that the younger cohort of women (70 to 71) with higher exposure to the reforms have a lower likelihood of falling below 150 or 200 percent of the poverty line. This trend is reversed for the older cohort (75 to 76) who see an increased probability of falling below 200 percent of the poverty line from RET elimination. This lends credence to the theory that a disproportionate impact of reduced benefits from early claiming fall on older women who do not experience a commensurate or countervailing increase in earnings that younger cohorts are able to exploit. Hence, it would appear that younger women can increase their labor supply over the longer period of their exposure to the reform thereby increasing their earnings and offsetting the impact of early claiming. Older women, on the other hand, experience labor supply increases over fewer time periods, since they are closer to retirement.  They are therefore less likely to offset reduced incomes from early claiming.

CONCLUSIONS

FN conclude that this analysis reveals important implications of RET elimination on poverty incidence for older age women, and suggest possible research extensions to include the influence of other sources of income, savings, and widowhood.

Read the working paper.

Read the research brief.

Contact MRRC.

[michiganretirementresearch]

Welfare of U.S. vets has declined over time, researchers find

As the face of America’s veterans change, policy, too, will adapt. In this 2015 working paper, “Declining Wealth and Work among Male Veterans in the Health and Retirement Study,” Alan Gustman, Thomas Steinmeier, and Nahid Tabatabai (GST) document the changing demographics of veterans and their preparedness for retirement. They discover something that the authors can only call “a reversal of fortunes” for these veterans:

“Veterans ages 51 to 56 in 1992 were better educated, healthier, wealthier, and more likely to be working than nonveterans. By 2010, veterans had lost their education advantage, were less healthy, less wealthy, and less likely to be working at ages 51 to 56 than nonveterans.”

The authors base their analysis on the Health and Retirement Study (HRS), using survey data from 2002 to 2010. They focus on males born from 1936 to 1959 who were ages 51 to 56 in the first year they were surveyed. (Only 1 to 3 percent of veterans in older cohorts were women, which leads to insufficient numbers for analysis.) Their sample includes veterans from the Vietnam War and the Gulf War, veterans who were not deployed in combat, and nonveterans.

In particular, the youngest cohort used in this study, the Mid Boomers, saw the advent of the all-volunteer military. Unfortunately, the armed forces had to endure some growing pains in the first few years of this program. Military pay was not at the competitive levels envisioned at the beginning of the all-volunteer force and administrative error on the Armed Forces Qualifications Test meant that many applicants who should have been rejected for military service were accepted from 1977 to 1980. Those would have been prime years for the Mid Boomers to enter the military. This gives one good reason to suspect the Mid Boomers cohort of veterans is very different from older cohorts who were drafted into the military.  They were also the hardest hit by the Great Recession (compared to earlier cohorts), which reduced retirement wealth for veterans and nonveterans alike.

Calculating wealth

In order to study preparedness for retirement, the authors take into account all financial assets including housing, liquid assets, social security and pension wealth. The authors take great care in constructing an individual measure for wealth in pensions by calculating pension wealth associated with each job reported in the HRS. Their method is detailed in their book and has resulted in values for pension wealth that are often used by fellow researchers.

One caveat the authors point out is that veteran’s pension is only available to those with 20 years or more of service. According to survey response, only about 6 percent of each cohort has 20 years of uninterrupted active service. For these respondents, military pension will be accurately computed. However, respondents with 20 years of service from two or more noncontiguous active spells, or a combination of active and reserve spells, may also be eligible, but will be treated as not having military pension in this analysis, leading to lower estimates for financial wealth than they actually hold.

Social Security wealth is calculated with the Social Security Administration’s ANYPIA program. ANYPIA uses an individual’s work history and dates from military service and reports an individual’s Primary Insurance Amount (PIA), with adjustments for military service. (An online version is available here should the reader choose to undertake this exercise with his/her own employment records.)

Financial wealth, as measured by PIA, Present value of potential SS benefits, and present value of all pensions, indicate that older veterans (born 1936 to 1947) were wealthier than nonveterans, but the trend is reversed for younger cohorts born after 1948. For Mid Boomers, present value of all Social Security payments is 20 percent lower than nonveterans. Similarly, employment rate at ages 52-56 is higher for veterans than nonveterans of earlier cohorts, but the trend is reversed by the Mid Boomer cohort.

Findings

For the older cohorts, observable differences in education, race, marital status and health explain most of the wealth gap. Even after controlling for the aforementioned aspects, however, Mid Boomer veterans had 10 percent less wealth than nonveterans.  If poor health and less opportunity for marriage is a result of military service, then the economic penalty of military service is even larger.

There are two explanations for this shortfall in wealth for Mid Boomer veterans. It could be a symptom of those with lower earnings potential self-selecting into the military, where the military experience itself does not penalize veterans. However, it could also be due to lack of economic opportunity for the all-volunteer force. If it is the latter, then expanding SS benefits for veterans may narrow some of this wealth gap. The authors hope that data on more recent cohorts of the all-volunteer force who were not subject to some of the earlier mistakes may shed more light on this issue, but are skeptical that the gap will disappear entirely in more recent cohorts.

Read the working paper

Read the research brief

 Contact MRRC.

[michiganretirementresearch]

How do return Mexican migrants interact with Social Security?

Under the 1986 Immigration and Reform Act (IRCA), nearly 2.7 million undocumented immigrants were granted legal, permanent residence in the United States. Nearly 70 percent of these immigrants were of Mexican origin, and nearly half will be 52 years of age or older as of 2012 and, therefore, part of the pool of potential Social Security beneficiaries. This generates a renewed interest in the impact of immigrants on the Social Security system.

A seldom researched category of immigrants is return migrants, whose numbers are increasing over time. Between 1995 and 2010, return migrants 65 years of age and older rose from 1.5 percent of the migrant population to 6 percent. This category of migrant is also most likely to be missing from most U.S. databases. Therefore, their effect on Social Security remains unexplored. In the 2015 working paper ”Social Security Contributions and Return Migration among Older Male Mexican Immigrants,” Emma Agulia and Alma Vega (AV) explore the demographic characteristics of older Mexican males who are characterized by return migration, and how this group interacts with the Social Security system during their time in the U.S.

Immigrants may pay Old Age, Survivors and Disability Insurance (OASDI) but never claim their benefits if they return to their home countries before becoming eligible. On the other hand, they may return to their home countries after qualifying for benefits and, therefore, transfer U.S. resources abroad. Post-IRCA, while a legal immigrant is entitled to collect benefits from wages made while he was undocumented, it is thought that few actually provide requisite documentation for these wages. This further complicates their ability to actually collect benefits even when given legal status. Consequently, in order to assess the impact of return migrants on the Social Security system, we must turn to empirical analysis.

AV use the Mexican Health and Aging Study (MHAS—closely modelled on the Health and Retirement Study), which collects panel data on Mexican adults older than 50 and oversamples regions with strong U.S. migration. LV’s sample consists of males in the 2003 and 2012 MHAS waves older than 50 who reported ever having been in the U.S.

Among this sample, 40 percent report having contributed to the Social Security system in the 2003 wave of the panel, and 32 percent in the 2012 wave. AV consider this an underestimate since some respondents may not have been aware that they were contributing. Meanwhile, only 5 percent received or expected to receive benefits. Most of the contributors were of older age and lived in the U.S. for, on average, twice as long as those who did not contribute—seven years versus 3.2 among noncontributors. They were also twice as likely to be U.S. citizens or permanent residents—15.3 percent versus 6.5 percent of noncontributors. AV corroborate these statistics with a simple regression analysis, which reveals that higher education, longer time of stay in the U.S., and legal residence are all strong predictors of the probability that a return migrant contributed to the Social Security system.

AV also explore how return migration may affect transitions out of work and into retirement. Relative to males who have never lived in the U.S., respondents who had lived in the U.S. for less than 10 years are less likely to be retired, while those who had lived in the U.S. for longer were more likely to report not working. This would suggest that the shorter span of migration leaves migrants less prepared for retirement.  

AV conclude on the cautionary note that their analysis cannot prove causality. However, it is the first work of its kind to investigate features of return migrants and how they relate to the U.S. Social Security system.  

Read the working paper.

Read the research brief.

Contact MRRC.

[michiganretirementresearch]

Researchers examine shortcomings of retirement preparedness metric

The Income Replacement Rate (IRR)—income immediately following retirement divided by income immediately preceding retirement—is the prevailing metric for economic preparation for retirement. Target IRRs are meant to serve as a simple rule of thumb for financial planning, with a 70 percent minimum IRR recommended by many financial advisors. However, while simple as a retirement planning tool, the IRR also has shortcomings.

The puzzle

Two-thirds of individual career trajectories do not proceed discretely from work to retirement, but instead include transitions between full-time work, part-time work, unemployment, unretirement, or disability. It is, therefore, difficult to establish what would constitute preretirement and post-retirement income for the calculation of an IRR. The traditional calculation of IRR also does not account for differences between single-earner and two-earner households, draw down of savings, changes in consumption paths with age and the impact of pre- and post-retirement taxes on income.

In their 2015 working paper “Measuring Economic Preparation for Retirement: Income vs. Consumption,” Michael D. Hurd and Susann Rohwedder (HR) construct several measures of the IRR to try to account for the above shortcomings. They compare these adjusted measures with a consumption-based measure of economic preparation developed in their earlier work. The latter measure assesses whether economic resources at retirement can support stable consumption till end of life, while making appropriate assumptions regarding impact of taxes, savings, possibility of health shocks, widowing, and uncertainty regarding time of death.    

Data and measures

For the measures of IRR, HR use data from the Health and Retirement Study (HRS) waves 2002-2008. They add data on Social Security benefits and Defined Benefit (DB) pensions from their earlier work (referenced above) and work with a subsample that was included in the Consumption and Activities Mail Survey (CAMS) so that they may construct their consumption-based measure for comparison. They first construct a standard IRR, dividing the sum of Social Security and DB income by average income prior to years 59 to 61. Their second measure, IRR- Fixed (IRRF), adds a fixed 4 percent of financial assets and IRAs to post-retirement income. Their third measure, IRR-Annuity (IRRA), adds annuitized income from financial assets and IRAs.

Results for singles

For single people, the standard IRR calculations predict that only 35 percent of individuals in the data are prepared for retirement with an IRR of over 0.7. Adding a fixed 4 percent income from assets and IRAs, the IRRF measure predicts that 46 percent of the single sample has a sufficient post retirement income as measured by a replacement rate of more than 0.7. The IRRA, which adds annuitized income from financial assets and IRAs, predicts that 54 percent of the sample has sufficient post retirement income. The consumption-based measure of economic preparedness predicts that 59 percent of the sample are adequately covered, and therefore relates closest with the IRRF and IRRA.

Results for couples

When considering couples, HR develop two estimates of IRRs. The first set of estimates considers each spouse as a single individual, and uses their individual preretirement income and post-retirement assets. The second measures treat the couple jointly and combine their preretirement incomes and post-retirement assets. With this joint income and asset measure, they construct the CIRR, CIRRF and CIRRA, the couple-equivalents of the IRR, IRRF, and IRRA respectively.

When treated as individuals, couples perform worse than the sample of singles. However, their replacement rates improve when their income and assets are considered jointly, making the number of couples with CIRR, CIRRF, and CIRRA over 0.7 virtually indistinguishable from the sample of singles.

Conclusions

This analysis of replacement rates suggests, therefore, that couples and singles are equally prepared for retirement. HR assert, however, that the actual data on post-retirement income and financial wealth shows that couples have an advantage over singles. The consumption-based measure of retirement preparation does accurately predict that a significantly higher proportion of the couples’ sample is adequately prepared for retirement than singles (80 percent vs. 59 percent). HR suggest, therefore, that the consumption-based measure may be preferable and the IRR measures may be less compelling in terms of their theoretical bases, as well as their empirical predictions.

It would appear that the financial planning industry has also arrived at a similar conclusion, with a shift from simple replacement-rate analyses toward providing retirement planning advice based on spending needs and priorities, as a consumption-based metric would do.

Read working paper

Read research brief

Contact MRRC.

[michiganretirementresearch]

ACA’s effect on labor-force participation among older workers still unclear

The Affordable Care Act (ACA) has faced considerable political resistance partly because of the possibility that it will reduce labor supply and increase retirement by expanding the availability of low-cost private insurance not tied to workforce participation.  The Congressional Budget Office has estimated that the ACA will reduce hours worked by 1.5 to 2 percent between 2017 and 2024, the equivalent of 2 to 2.5 million fewer workers.

This concern has its underpinnings in a comprehensive literature that demonstrates that the availability of health insurance options not tied to continued employment increases the likelihood of retirement in working adults. A related strand of research has also shown that the expansion of public insurance (Medicaid) coincides with reductions in labor supply, while reductions in Medicaid eligibility increase labor supply, with effects being concentrated among older adults.

The implementation of the ACA in January 2014 has had the dual effect of increasing Medicaid eligibility and reducing the cost of private insurance options. Some states also opted into the expansion of federally funded Medicaid coverage for households with income below 133 percent of the federal poverty level.  As of July 2015, 30 states and the District of Columbia have opted into this expansion.

In addition, the ACA introduces reforms in the private insurance market that significantly reduce the effective price faced by older adults who are not eligible for employer-based insurance. These reforms include prohibitions on private insurers denying coverage based on pre-existing conditions; creation of insurance exchanges that facilitate comparison across providers; subsidies on insurance premiums for low-income individuals; and, most importantly, limits on how much premiums can vary with age.

In theory, therefore, the ACA has created incentives for a reduction in labor supply amongst older adults. In their 2015 “The Effect of Health Reform on Retirement,” Helen Levy, Thomas C. Buchmueller, and Sayeh Nikpay (LBN) assess whether these labor market effects have been observed empirically in the 18 months following the enactment of the ACA.

Data, model, and findings

The authors employ data from the monthly Current Population Survey (CPS) between January 2005 and June 2015 to measure changes over time in the stock of retired people 55 to 64 years old.  In addition, they match individual observations across months to evaluate monthly flows in and out of retirement. The authors also use the 55 to 64 subsample of the 2012 Health and Retirement Study (HRS) for supplemental analyses.

LBN first use the 2012 HRS to establish the size of the pool of workers who will be most affected by ACA, insofar as they are covered by employer-provided plans which they will lose upon retirement. This amounts to roughly 40 percent of the 55- to 64-year-old subsample.

LBN then use a multivariate regression model to assess whether the onset of the ACA in 2014 and the increase in Medicaid coverage in expansion states relative to nonexpansion states result in changes in retirement trends. They find no significant difference in overall retirement trends following the enactment of the ACA, and no significant difference in trends between expansion and nonexpansion states that may have been attributable to Medicaid expansion. The CPS data would therefore suggest that ACA provisions overall, and particularly the provisions allowing for Medicaid expansion, have not affected retirement trends in the 18 months since ACA enactment.

The authors test a similar model on CPS data matched from one month to the next to assess the impact of ACA on monthly and annual transitions from work to retirement. However, they find that since they are working with a smaller dataset, the relationship between ACA and work transition is imprecisely measured and therefore their results are inconclusive.    

LBN also explore the possibility that older workers switch from full-time to part-time work after ACA enactment, instead of retiring outright. They find no consistent trends in part-time work overall, and no significant change in trends between expansion and nonexpansion states post 2014. Parsing the data by education, they do find an increase in part-time work among workers without a high school diploma in expansion states in late 2014 and early 2015 that is not mirrored in nonexpansion states. However, they need more data to establish that this is trend will persist going forward.

LBN acknowledge, however, that difficulties in initial enrollment for ACA and uncertainties over the legal challenges to its enactment may have depressed labor supply response over the period being considered. The authors hypothesized trends of reductions in labor supply or greater switching to part-time work may emerge in the future, as more data is collected. However, over the period of January 2014-June 2015, ACA appears to have not affected labor supply in any statistically meaningful ways.          

Read working paper

Read research brief

Contact MRRC.

[michiganretirementresearch]

Researcher Q&A Part 2: Olivia S. Mitchell looks at why people do or don’t buy Long-Term Care Insurance

Here’s the second part of our discussion with MRRC researcher Olivia S. Mitchell on her 2015 project with Daniel Gottlieb, “Narrow Framing and Long-Term Care Insurance.”

This has been edited for length and clarity.

I think a lot of our readers are not familiar with the long lead times involved with large scale surveys such as the Health and Retirement Study. At the time when your team designed the survey question, it was about 2010? The question was fielded in HRS 2012.

So you have to do it two years before.

…and the results don’t come until 2014.

Two plus years later.

So what is it like waiting four years to see if you were right?

I like to tell the story of former colleague of mine who was a rocket scientist. He came home one day from work and our kids were playing together, and he said, “Today I shot a rocket up for my son and my grandchildren to be able to study because it’s going to the far reaches of the universe.” So what’s four years, you know?

You do have to think ahead. I think that perception has really guided us since the beginning of the HRS. I remember the first meeting I ever went to about the HRS, before it was born. It was in 1980, and that was 12 years before the first wave was ever fielded. Here we are, in our 20s, looking ahead to the 2016 wave… you just have to try to design questions that you think are fundamental to the field. And then maybe someday later, people will be able to use them in ever new waves.

Before the results were released, was your team convinced that this approach (see previous post) would work?

No…The conventional approach in economics is that people are rational, people make the best decision they can, given what they have at their disposal. I still think that that’s mostly true. However, we do see this persistent, unexplained behavior around why people don’t buy insurance.

Another area I work a lot on is annuities, which are purchases that you make, say, at retirement and then they pay you a lifetime income stream. People don’t buy annuities either, and that’s another big puzzle. So there are two or three big puzzles that come up in the insurance context. If you talk to average people, people on the street [and ask], “Why don’t you think about buying annuity at retirement,” they’ll say, “I have to pay a bunch of money up front to the insurance company, and I might get hit by a bus the next day and I won’t get any benefit back.” I say, yes, but you will that peace of mind. It’s the exact same problem, the notion that you’re very reluctant to plop down a bunch of money even if it is buying lifetime security.

If people view buying insurance as a normal investment vehicle, then what’s wrong with choosing a portfolio of investment vehicles to hedge against lifetime expenditures?  What is the difference, is it the peace of mind?

It depends on the product. If you’re talking about annuities, an annuity is the only financial instrument that helps protect you against outliving your assets. There’s nothing else like it. Now, if you’re Bill Gates, you don’t need to have a protection against outliving your assets. And if you’re very poor, then you don’t have any assets anyway. It’s people in the middle that could possibly run out of money if they outlive their assets. The annuity is the only financial product that can provide at least a minimal income stream, guaranteeing minimal consumption. There is no other asset that does that …How many people have ever thought about their health insurance as an investment? It’s insurance, it’s not an investment. How many people think about their automobile insurance as an investment? It’s not an investment, it’s protection. You put down money and in exchange you get protection.

In the case of an annuity, you put down money and in exchange, you get an income stream for as long as you live. In the case of LTCI, you plop down some money and, if you end up going into a nursing home for whatever ungodly price it might be, you have protection. I think it’s important that you try to educate people about the difference between insurance and investments.

I grant you that there are some products that try to blur the distinction between them, and I think that’s one reason a lot of people are suspicious about insurance, because they don’t really understand the product. Frequently they’re very complicated. I’d like to see a simplification of insurance products such that there would be a standardized template that could be used to compare insurance products... I do think that the insurance industry can do a lot more to make clearer what it is that [purchasers are] getting and what it is that they’re giving you in exchange for the fees. This has been a long-term issue and I hope that it will start to become clearer.

You’ve been doing retirement research for a long time, how much of this gap of LTCI coverage was foreseen in the ’70s and ’80s?

Longevity has been increasing, that certainly is true and has been going on for a long time…I suppose one of the other pieces of the puzzle is that in the old days, the families would take care of the elderly, but that’s not happening anymore in many societies. Especially with complex families—broken marriages, divorce, remarriage, whatever—it’s a lot less clear who’s going help take care of Mom.  That’s another reason why LTCI should be on people’s agendas.

Read the working paper      

Read the research brief

Contact the MRRC

[michiganretirementresearch]

Researcher Q&A, Part 1: Olivia S. Mitchell looks at why people do or don’t buy Long-Term Care Insurance

We talked to MRRC researcher Olivia S. Mitchell about her 2015 project with Daniel Gottlieb, “Narrow Framing and Long-Term Care Insurance.”

This two-part interview has been edited for length and clarity.

What led you to look at whether people buy long-term care insurance (LTCI)?

The background to the project came from the thinking that I and some of my colleagues had done around insurance. That is, why don’t people buy insurance when they are facing very big, potentially catastrophic costs? This is a pervasive finding across a whole range of different topics, but long-term care is one of them.

My own mother recently passed away, and she was in a LTC facility and it was about $85,000 a year—and she was there for a long time. When one thinks about these big expenses, it makes sense to think about trying to buy insurance against these... The problem that we find is that many people don’t buy LTCI, notwithstanding the big expense and high probability of needing care.

My colleague [Daniel Gottlieb] and I started thinking about why people might not buy LTCI. There are many arguments that people have made previously. One is that if you spend all your money, then the government will support you through the Medicaid plan and pay for your LTC—maybe not in such a good place as you might otherwise be in, but many nursing homes have a restricted number of spots for people who are on Medicaid. So if you run out of money, you can throw yourself on the mercy of the government. That’s one possibility.

Another possibility people have suggested is that if you have LTCI, maybe your children won’t take good care of you because they want you to leave your own home and get in the nursing home... Those are all arguments that we have heard about before and other people have studied.

What we were more interested in is the fact that when you buy LTCI, it’s expensive, it costs you $3,000 to $4,000 a year depending on the plan that you have. People see that expense as a waste of money because they won’t need the care until much longer into the future. They’re focusing on the near-term cost and not understanding the long-term benefit of having coverage.

How did you look at the question of why people don’t buy LTCI?

What we had was a module for the HRS (Health and Retirement Study) that allowed us to try to evaluate some traditional questions that had been developed by psychologists. Tversky and Kahneman are two very well-known psychologists [who] have focused on how the way you frame questions to people influences the way they answer them.

We had two questions... straight out of the Tversky and Kahneman approach… two hypotheticals. One was, imagine there’s going to be an epidemic expected to kill 600 people. There are two programs proposed. If Program A is adopted, 300 people will be saved. If Program B is adopted, there’s a 50-50 chance that either 600 people will be saved or none will be saved... We showed people that and asked which they would prefer…

The truth of the matter is that, if you know your statistics, those two programs are exactly identical, but for some reason people seem to latch on to Program A, which says that 300 people will be saved. They feel that that’s much more secure than a 50-50 chance.

We asked them that question, then the second question is basically the same construction, but instead of framing it in terms of people who would be saved, it’s framed in terms of people who will die. So we say again, there’s an epidemic and 600 people may be killed. If we adopt Program A, 300 people will die—which is just the inverse of the previous one. If we [adopt Program] B, then there’s a 50-50 chance that none will die or all will die. Again, we let people choose which one they prefer. The statistics are identical in both experiments, in both wordings, but when you show people this second formulation, more people are likely to shy away from the statement that says 300 people will die because they don’t like that notion.

We randomize the order of these presentations, so some people got question one first, some people got question two first. Then what we said was anyone who answers the questions similarly in the two cases, they understand that this is really the same problem. They are not being unduly influenced by the notion of this big loss, “300 people will die.” Whereas anyone that answers them inconsistently is looking just at the framing of the questions and being unduly influenced by the words “will die or will be saved.”

Using that, we said that if people were narrow framers, and they’re also loss averse— they don’t like the idea of people dying, which most of us don’t—people would choose the safe option when we suggested “300 will be saved,” but they’ll choose the risky presentations when we say “300 will die.” Those are the inconsistent people. Those are the folks that we said were narrow framers.

In a separate part of the question, we asked them whether they had LTCI—this was part of the core HRS, actually. Then essentially what we did was a simple correlation between whether you were a narrow framer—that is your answers changed in the two presentations—and whether you had LTCI. It turned out that the people who were narrow framers were much less likely to have LTCI.

Our interview with Dr. Mitchell continues in the next post.

[michiganretirementresearch]

Researchers find that Social Security reform decreases Medicare spending

The existing literatures on Social Security and Medicare explore each of these significant allocations of federal budget in isolation. There has been no work to date on the manner in which these programs interact with one another. Did Social Security Amendment of 1983 affect governmental Medicare spending? In their 2015 working paper “Medicare Expenditures, Social Security Reform, and the Labor Force Participation of Older Americans,” Yuanyuan Deng and Hugo Benitez-Silva [DBS] attempt to address this question by estimating the cost savings that Medicare derives from changes to the Social Security Old Age benefits system.  The authors contend that a comprehensive evaluation of the impact of Social Security reform should factor in its spillover benefits from reduced Medicare spending on older adults.  

Social Security reforms fall under two key provisions: increases in the Full Retirement Age (FRA) and the Delayed Retirement Credit (DRC), both of which have been shown to decrease Social Security benefit receipts. In order to offset this reduction in post-retirement income, the Labor Force Participation Rate has increased at older ages. If an older individual is covered by employer-provided health insurance, this delay in retirement will result in lower Medicare expenditures since private insurance will operate as the primary payer for health care expenditures, and Medicare will be a secondary payer. Moreover, the availability of employer-provided health insurance may also delay the decision to enroll in Medicare, resulting in reductions in Medicare expenditures overall. These are the channels that DBS set out to investigate.

The existing evidence on determinants of Medicare expenditures suggests that the availability of supplemental insurance actually increases Medicare spending, which runs counter to DBS’ hypotheses. However, this literature does not deal exclusively with the older demographic and fails to account for cost savings that emerge from having Medicare as a secondary payer and from delayed enrollment into Medicare, making DBS’ paper the first of its kind.

DATA

The authors use data from the Cost and Use files of the Medicare Current Beneficiary Survey (MCBS) over the period 1999-2010, with an estimation sample comprising elderly Medicare beneficiaries around retirement age (65 years and older). Since there are differences in the timing of enactment of DRC and FRA rules, some rules can affect one age cohort in the data and not another. By identifying the beneficiary’s birth year, the authors establish a beneficiary’s age cohort and the precise DRC and FRA rules to which the beneficiary was subject. The authors can now compare across different age cohorts to parse how outcomes for one cohort differ from another, based on changes in these DRC and FRA rules between cohorts.

FINDINGS

The results of their empirical estimation reveal that workers who are covered by current employer-provided health insurance generate 26.4 percent lower Medicare expenditures relative to workers who are not covered. Moreover, spouses of workers who are covered generate an additional 35.2 percent savings in Medicare expenditures relative to spouses who are not covered by employer-provided insurance. After accounting for the effect of working and having employer-provided health insurance, individual DRC and FRA provisions continue to show a negative impact on Medicare expenditures and total health expenditures. The authors attribute this to cohort effects, arising possibly from increases in health investment between cohorts over time.  

DBS estimate the average dollar reduction in Medicare expenditures for individuals with Medicare as a secondary payer at $4,568 per person, per year, for total savings of more than $2 billion across the system. Additionally, they estimate a reduction in Medicare expenditures of $838 million deriving from an increase in the number of individuals who have zero Medicare expenditures from Medicare being a secondary payer, for a yearly net saving of $2.89 billion (0.65 percent of total Medicare expenditures in 2010).

DBS further find that labor-force participation and employer-provided health insurance are significantly correlated with delayed enrollment in Medicare, which generates additional Medicare savings for each year that enrollment is delayed. Using comprehensive estimates for Medicare cost savings specific to each age group, survey year, and number of years delayed, DBS estimate an additional $333.67 million cost savings in any given year from delayed enrollment.

DBS conclude that Social Security Amendment of 1983 generated significant reductions in Medicare spending through labor supply and health insurance coverage. Future research promises to extend this analysis to the impact of Social Security reforms on disability benefits as they relate to Medicare spending.

Read the working paper.

Read the research brief.              

Contact MRRC.