Social Security and Financial Security at Older Ages

Introduction

The age-based component of Social Security—Old-Age and Survivors Insurance (OASI)—is a major source of income for most older Americans. For the insured worker, benefit eligibility begins at age 62; and by deferring the claim (up to age 70), the beneficiary is entitled to actuarial increases in the monthly benefit amount. Once claimed, monthly benefits continue, with annual cost-of-living adjustments, for the remaining years of life. It would be hard to overstate the importance of this annuitized income stream to people's financial security at older ages. There are nearly 54 million current OASI beneficiaries, including retired workers and their dependents and survivors (Social Security Administration 2019, Table 2).

Although the core functions of Social Security remain largely unchanged, the program operates in an environment of continually changing demographics, health trends, longevity, labor markets, economic conditions, government finances, household finances, and related public and private programs. The dynamic evolution of these influences makes the ongoing monitoring and evaluation of Social Security policy and its implications an important subject of research attention and the explicit focus of the Retirement Research Center (RRC) at the National Bureau of Economic Research (NBER). The NBER RRC has been active since 2003 and operates through a cooperative agreement with the Social Security Administration. This article highlights key findings from the last 5 years of the NBER RRC's research. Findings from the NBER's companion Disability Research Center are described in a separate article in this issue of the Social Security Bulletin.

Work, Retirement, and Claiming Social Security Benefits

A number of RRC studies address decisions about work at older ages, retirement, and Social Security claiming. Many factors influence such decisions, including the worker's health, family circumstances, health insurance availability, assets, employer-provided pension coverage, earnings, and local and national labor market conditions, as well as Social Security policy. Many Social Security provisions feature implicit financial incentives to retire or claim earlier or later, such as the ages of eligibility for benefits, the treatment of earnings and benefits for those who continue or return to work, the benefit adjustment formulas that apply to claiming at different ages, the benefit amounts, and the tax treatment of benefits.

Shoven and Slavov (2014) track the changes in Social Security's implicit financial incentives over time. They look first at policy changes, such as the phased increase in the delayed retirement credit from 3 percent per year after the full retirement age (FRA) for workers born in 1924 or earlier to 8 percent per year for workers born in 1960 or later, and the relaxation of restrictions on the timing of claims for nonearning spouse benefits. They find that the rule changes increased the gains from delayed claiming by 1–2 percentage points for singles, 5–6 percentage points for two-earner couples, and 2–4 percentage points for one-earner couples, and that most of those increases are attributable to the rise in the delayed retirement credit. The authors also find that longer life expectancies and real interest rate declines over the study period further increased the financial reward for delayed claiming.

Börsch-Supan and Coile (forthcoming) summarize a cross-national project documenting changes in social security policy in the United States and 11 other higher-income countries over the last three decades and review how those reforms affected financial incentives to retire at different ages. They find that most but not all of the reforms lessened the effective tax on work at older ages. Such changes in work incentives in the United States and elsewhere partially explain why labor force participation at older ages has increased since 1990, reversing a previous trend toward younger retirement.

Other RRC projects analyze the effect of specific reforms on behavior, often drawing, like Börsch-Supan and Coile, from the wider scope of reforms implemented in other countries. Four recent studies find a strong relationship between changes in benefit eligibility ages and labor market behavior. Lalive and Staubli (2014) analyze how women's work, retirement, and claiming behavior changed in response to a Swiss reform that increased the FRA from 62 to 64. They find that a 1-year increase in the FRA delays labor-market exit by 7.9 months and social security claiming by 6.6 months. In a follow-up study, Lalive, Magesan, and Staubli (2017) find that some of that delay is caused by “rule of thumb” behavior in which workers accept the FRA passively, and some of the delay results from changes in financial incentives caused by the reform. Manoli and Weber (2016) find that a 1-year increase in Austria's early retirement age leads to a 0.4-year increase in the average age of job exit and a 0.5-year increase in the average pension claiming age. Manoli and Weber (2018) focus on developing firm-level data for Austria, but also report preliminary evidence that firms employ increasing numbers of men aged 60–63 and women aged 55–58 as increases in the eligibility age for early retirement benefits phase in.

Two recent studies look at changes in the formula for adjusting retirement benefits according to claiming age. Brinch, Vestad, and Zweimüller (2016) analyze a 2011 reform in Norway that allows pension claiming anytime from age 62 through 75, with actuarially neutral pension adjustments for claiming at different ages. Cumulatively, the reforms replaced the old system's implicit tax on earnings after age 62 with incentives to retain employment. The study finds that registered employment among those aged 62 to 65 increased 26 percent, earnings increased 15 percent, and the rate of retirement at age 62 dropped substantially from prereform levels. Lalive and Staubli (2016) analyze a Swiss reform that doubled the benefit reduction for early retirement from 3.4 percent to 6.8 percent annually. They find that the adjusted formula led to a 4–5 month delay in pension claiming, but no change in labor force exit in that country.

A recent three-phase RRC project explores the effect of the Social Security earnings test on labor market behavior in the United States. Gelber, Jones, and Sacks (2014) find that the earnings test not only creates bunching around the earnings limits defined by the formula but also leads some individuals to forgo work altogether. Gelber and others (2017) reinforce those conclusions, finding that the employment rate of workers subject to the earnings test decreases significantly relative to that of other workers. For people aged 63–64, for example, the earnings test reduces the employment rate by at least 3.7 percentage points. Gelber, Jones, and Sacks (2019) present evidence of earnings-test frictions, wherein some individuals continue to behave as if they are constrained by the earnings test even when they are no longer affected. Specifically, during the period from 1990 to 1999, when the earnings test was imposed at ages 62–69 but not at ages 70 or above, there was still a modest bunching of earnings at ages 70 and 71, as if the earnings limit still applied. Even though the reductions in short-term Social Security benefits caused by the earnings are largely or completely offset by increases in Social Security benefits later, these results imply that eliminating the earnings test could increase work among older people.

Some state pension systems impose another form of earnings test, limiting the number of hours beneficiaries can work and still receive a state pension. Fitzpatrick (2019) finds that raising the maximum number of work-hours allowed increases part-time work among retirees without reducing retirement benefits collected. As such, these policies appear to be binding on the employment decisions of some state pension beneficiaries, just as the earnings test affects some Social Security beneficiaries.

The studies described so far emphasize how Social Security policy affects work, retirement, and benefit claiming. Of course, other influences affect behavior too. For example, ease of access to health insurance before Medicare eligibility at age 65 may influence early-retirement decisions. Coe and Goda (2014) analyze how state-level reforms that prohibit insurers from rejecting applicants (guaranteed-issue requirements) or charging different rates based on health (community-rating regulations) affect early retirement. They find that these regulations, when applied in the individual (nongroup) market, substantially increase labor force withdrawal. At age 63, the monthly probability of retirement increases by 2.2 percentage points, or nearly double the rate that would occur without the reforms. For workers in fair or poor health, who are most likely gain access to the individual market through these regulatory changes, the effects are even greater. Among individuals predicted to be more sensitive to health policy regulations, access to health insurance accelerates retirement by 1–2 years.

Banks and Coe (2017) look at how the regulatory reforms and subsidies in the Patient Protection and Affordable Care Act (ACA) affected retirement expectations. Using survey data, they find that 15 percent of respondents indicated that they would be more likely to retire early. Heim, Lurie, and Simon (2016) compare retirement behavior in states that expanded Medicaid eligibility after enactment of the ACA with states that did not expand eligibility. Although they find no effect of Medicaid expansion on retirement behavior for the population generally, they detect small effects among women and single individuals who had access to employer-provided health insurance while working.

The causal link between health insurance and Social Security can work in the reverse direction as well. Fitzpatrick and Moore (2016) examine whether the increase in the Social Security FRA from 65 (for workers born in 1937 or earlier) to 66 (for workers born 1943–1954) delayed people's Medicare enrollment. The authors find that at age 65, the rate of Medicare participation is 2.5 percentage points lower for cohorts with an FRA of 66 than it is for cohorts with an FRA of 65. By age 66, there is again no difference in Medicare enrollment rates. This suggests that some people affected by the older Social Security FRA delay Medicare enrollment from age 65 to 66, but not beyond age 66.

Several RRC studies look at the psychology of Social Security claiming, and how framing the claiming options influences behavior. All of these studies are based on experimental surveys and interventions that alter and compare the framing, presentation, informational content, or messaging about Social Security claiming options. Greenberg and others (2017) find that Social Security claiming plans are sensitive to one's longevity expectations and how a person is asked to think about them. For example, if respondents were asked to think about someone they knew who had lived long into retirement, they indicated an intended Social Security claiming age that was 9 months older, on average, than the average age reported by respondents who were not. Similarly, respondents who were asked to focus on dying early in retirement indicated claiming intentions that were 9 months younger than the average age indicated by respondents with no such prompt. In follow-up work, Greenberg and others (2018) look at the effect of messages designed to promote self-reflection and to help people make reasoned judgments about Social Security claiming age. They test numerous messages, most of which significantly delay claiming intentions toward older average claiming ages, ranging between 5 months and 10 months later than the ages reported by respondents who do not receive such messages.

Two studies look at what happens to claiming expectations when respondents are given additional information on the cumulative payouts that would be made from Social Security under alternative scenarios. The findings differ. Shu, Payne, and Sagara (2014) compare the effects of two different tabulations of Social Security benefits, one showing monthly benefit amounts and the other showing cumulative lifetime payouts under various claiming-age and lifespan assumptions. They find that the lifetime-payout presentation leads respondents to favor earlier claiming, on average, over later claiming. Modrek, Reed, and Carstensen (2016) also test the effect of providing a table that compares lifetime benefits by claiming age and longevity, along with a modified SSA message highlighting the possible advantages of delaying claiming. They find that women seeing this table delay their expected Social Security claiming age by 1.3 years on average. The findings from this work highlight the very significant effect that even subtle variations in the framing of claiming options can have on planned retirement and claiming behavior.

Shu and Payne (2013) find a relationship between claiming-age expectations and longevity expectations. For example, respondents expecting to claim Social Security at or before age 65 express an average probability of living to age 85 of 41 percent, compared with 50 percent for those expecting to claim at age 67 or later. Colby, Shu, and Payne (2017) likewise find that individuals who believe they will live longer choose to claim Social Security later. The authors also test whether people's claiming expectations differ if they are presented with a list of prudent financial planning goals (such as acquiring wealth and paying off debts) or, alternatively, a list of enjoyable retirement activities (such as travel or hobbies). They find little effect of these messages on claiming expectations.

Payne and others (2015) find that loss aversion, which varies widely across individuals, is highly predictive of financial preferences, including Social Security claiming age. The authors measure loss aversion by asking participants to choose among a series of gambling choices and find that individuals with high loss-aversion scores indicate an average expected claiming age that is about 6 months younger than that of individuals with low loss-aversion scores. In other words, those with greater loss aversion appear, when considering delayed claiming, to weigh the loss of early benefit payments more heavily than the gain they could achieve from higher monthly benefit levels.

Continuing to work at retirement-eligible ages can contribute to the well-being of individuals and families, the financial health of the Social Security system, and the economy more generally. In addition, delayed claiming—by raising the annuitized payment stream from Social Security—can further improve household financial well-being as age advances. Given these potential advantages, an important goal of the RRC is to understand the decisions people make about work and claiming, and how they are affected by Social Security policy and other factors. Recent RRC research reveals how significantly the structure of Social Security policy, and the presentation of claiming options, can affect what people do.

Health and Financial Well-Being

A second area addressed by RRC research is the finances and the health of the aged. Although changes to Social Security policy have been relatively modest, the economic and demographic environment surrounding Social Security has changed considerably in ways that influence people's broader financial well-being. The growth of defined contribution 401(k)-type savings plans, and the associated decline in traditional defined benefit pension income, is one important trend. Another is the rise in labor force participation by women, such that retiring couples today are more likely to have substantial dual-income histories.

The lifetime courses of a worker's income, assets, and health are inextricably related. The strength and persistence of that interrelationship weaves through many RRC studies. For example, Chetty and others (2016) document the wide variation in life expectancy by income and geography. They find that the gap in life expectancy between the richest 1 percent and the poorest 1 percent of individuals is 14.6 years for men and 10.1 years for women. The gap is also widening. Between 2001 and 2014, life expectancy increased by 2.5 years for those with higher incomes but remained stagnant for those with the lowest incomes. Life expectancy among lower-income households also varies significantly across geographic regions. Abraham and others (2014) find that spatial variation in mortality is three times greater for those with annual incomes between $10,000 and $25,000 than for those with annual incomes of $75,000 or higher. As an illustration, mortality rates for low-income individuals vary from 4,800 deaths per 100,000 in Yuma, Arizona to 15,000 deaths per 100,000 in Vincennes, Indiana. Low-income mortality and the slope of the mortality-income gradient are strongly correlated with local health risk behaviors (smoking and obesity) and racial composition but are only weakly correlated with measures of health care access.

Poterba, Venti, and Wise (2013) find a very strong relationship between education and financial well-being. Having a high school diploma increases the expected balance in a retirement plan by more than $50,000 relative to not having a diploma, and having a college or postcollege degree increases expected account accumulations by almost $250,000. Follow-up work in Poterba, Venti, and Wise (2018) finds that people with less education and lower lifetime earnings are more likely to have low wealth at retirement. Using a threshold of $100,000 in total assets to define low wealth, the authors find that 45 percent of couples in the lowest quintile of lifetime earnings have low wealth at retirement, while only 7 percent of couples in the highest lifetime-earnings quintile have low wealth. Similarly, 51 percent of couples without a high school diploma have low wealth at retirement, while only 6 percent of those with a college degree do.

Venti and Wise (2014) find that enrollment in Social Security Disability Insurance is more than six times higher for people with less than a high school diploma than it is for people with a college degree or more. People in the latter group are more than 25 percentage points less likely to claim OASI benefits early than are people with less than a high school diploma. Education is associated with better health and higher employment, earnings, and savings.

Poterba, Venti, and Wise (2017) analyze data from the University of Michigan's Health and Retirement Study and find that median assets do not change significantly as people age. For example, 70 percent of respondents born during 1931–1941 who held less than $50,000 in total assets when last surveyed before death also had held less than $50,000 in assets when first surveyed. Among respondents born in 1923 or earlier who held less than $50,000 in assets when last surveyed before death, 52 percent also had held less than $50,000 in assets when first surveyed. Although the typical asset trajectory is relatively flat over time, people do exhibit asset declines in connection with important medical events or disruptions in family composition. Poterba and Venti (2017) find that although some diagnoses do not substantially affect people's assets, strokes are associated with average declines in net worth of $4,682 for low-wealth individuals and of $59,290 for wealthier individuals. New diagnoses of lung disease are associated with declines in net worth of $9,986 and $84,959 for low- and high-wealth individuals, respectively.

Several RRC projects focus on the financial circumstances of lower-income households. For example, Baugh and Wang (2018) analyze how the once-a-month timing of Social Security income affects household finances between payments. They find that during months with a 35-day pay cycle (those having a fifth Wednesday) rather than the more common 28-day pay cycle, beneficiary households experience 9 percent more overdrafts, 6 percent more bounced checks, and 12 percent more online payday loans per day. Also, a mismatch of 1 additional week between the benefit payment and the due date for major bills results in 13 percent more overdrafts, 42 percent more bounced checks, and 37 percent more online payday loans. However, Giambra, Hastings, and Shah (2017) find no effect of the once-a-month timing of benefit payments on health outcomes.

Meyer and others (2018) find that many household surveys inadequately account for in-kind transfers and contain errors in earnings reports, transfer reports, and assets. When corrected, the number of households in extreme poverty is much lower than the raw data suggest.

Xu and others (2015) examine how the Big Five personality traits—conscientiousness, emotional stability/neuroticism, extraversion, agreeableness, and openness to experience—relate to financial distress. They find that conscientiousness reduces the likelihood of missing utility bills, losing phone service, missing rent or mortgage payments, being insolvent, worrying about no food, and being on welfare. Neuroticism raises the likelihood of these distress indicators. Xu, Brown, and Roberts (2016) find that the personality traits of conscientiousness and openness to experience are positively associated, and agreeableness is negatively associated, with nonpension wealth at older ages.

RRC research also considers the well-being of other groups such as widowed individuals, whose Social Security benefits are affected by special policy rules, and immigrants, whose employment histories differ from native-born beneficiaries. Fadlon, Ramnath, and Tong (2017; 2018) analyze how the death of a spouse affects household income, comparing the effects on widows just below and just above the age-60 eligibility threshold for Social Security survivors' benefits. They find that eligibility for survivors' benefits increases the widow's net annual income by nearly $5,000, or more than 11 percent. Benefit eligibility also enables widows to meaningfully decrease their labor supply.

Two projects compare the labor market experience and earnings histories of native, legal immigrant, and undocumented immigrant workers. Borjas (2017b) finds that the employment rate of undocumented men rose dramatically from 1994 to 2014. The probability of employment is far higher for undocumented immigrant men than for legal immigrant men, which in turn is higher than for native men. The probability that undocumented immigrant women work is lower than the probability that legal immigrant women work, which in turn is lower than the probability that native women work. Borjas (2017a) finds that the age-earnings profiles of undocumented workers lie far below those of legal immigrants and native workers. However, about half of the gap disappears after adjusting for other socioeconomic characteristics. The wage gap between observationally equivalent undocumented and legal immigrants fell from 10 percent to 4 percent between 2005 and 2014.

Several recent RRC studies look at wealth dynamics among retirees in other countries. Ljunge, Lockwood, and Manoli (2013; 2014) analyze how a person's financial circumstances change following divorce or widowhood in Sweden. They find that wealth is not drawn rapidly down during years with no major change in household composition. In fact, wealth increases during many such periods, especially during the early years of retirement, up to ages in the early 70s. Both divorce and spousal death, however, lead to large declines in household assets, even if assets are measured on a per-person basis. Böheim and Leoni (2016) find that aged people face greater income-poverty risks in the United States than in the majority of European countries, particularly among advanced-age women. The gender gap in poverty rates, which exists with different sizes in virtually all Organisation for Economic Co-Operation and Development countries, is particularly pronounced in the United States. Milligan and Wise (2014) find that the expansion of Canada's public pension system over the last 50 years has coincided with large improvements in elderly living standards, as measured by either income or consumption. For people in their 70s, the authors estimate that the 2010 system reduces income poverty by 88 percent relative to the 1960 system.

Social Security, combined with Supplemental Security Income for very low-income households, provides a solid base of annuitized income and financial security through later life. For a sizable share of households, it is their only resource. Evident throughout RRC research is the strength of the relationship between education, income, wealth, health, functional ability, Disability Insurance enrollment, and mortality; and how individuals struggling in any one of these domains are more likely to be struggling in others as well.

Retirement Saving

Accumulated savings in 401(k) plans and similar retirement accounts are an additional resource supporting the postretirement financial needs of many households. Individuals manage and largely fund such plans themselves. Some planholders may get matching contributions and guidance from employers, but in general, people make their own decisions about how much to save and how to invest their savings. Thus, a third major area of RRC research focuses on the determinants of retirement saving.

Behavioral tendencies and biases clearly influence how much people save. For example, Brown and Previtero (2014) look at procrastination, which is a manifestation of a behavioral phenomenon called present bias. The authors identify individuals as procrastinators if they wait until the last day of their health-insurance open enrollment period to choose their plan. When offered a supplementary retirement savings plan, procrastinators were 2.4 percentage points less likely to participate than nonprocrastinators, and they took 44–85 days longer to sign up, contributed less, and were less likely to annuitize their plan distributions.

Goda and others (2015) estimate the effects of present bias and exponential-growth bias (the tendency to underestimate the compounding of investment earnings over time) on retirement saving decisions. People with stronger exponential-growth bias, as estimated from survey responses, tend to save less. Present bias is the tendency for people to value benefits in the present over the future in a dynamically inconsistent way (that is, they act impatiently in the present but want their future selves to act more patiently). People with a stronger present bias also save less. The authors estimate that retirement wealth could increase by as much as 70 percent if present bias and exponential-growth bias were eliminated. In a follow-up study, Goda and others (2017) find that a one-standard-deviation increase in present bias corresponds to a one-third lower likelihood of making an active saving choice and a one-quarter lower likelihood of maximizing plan contributions. They also look at the effects of financial literacy on saving, as measured by a series of questions that test people's understanding of inflation, diversification, compound interest, mortgage payments, and bond prices. They find that a one-standard-deviation increase in financial literacy corresponds to an 18 percent increase in the likelihood of maximizing plan contributions.

One approach to overcoming procrastination and other adverse behavioral influences is through automatic plan enrollment with defaults for the contribution rate, investment allocation, and other features, from which a participant may actively opt out. These features have been shown to substantially increase participation, contributions, and asset accumulations. Beshears and others (2015) confirm that automatic enrollment leads to markedly higher plan participation rates. For employees who accrue 5 years with their employer, average plan balances as a fraction of starting salary are 10 percentage points higher under automatic enrollment than they are without automatic enrollment. The study also finds that employees subject to automatic enrollment have higher rollovers and cash withdrawals, but these differences only slightly counteract their greater 401(k) accumulations. Beshears and others (2016) examine whether the increased saving induced by automatic enrollment is offset by borrowing outside the plan. They find that 4½ years after hire, the effect of automatic enrollment on cumulative contributions as a percentage of first-year salary is 6 percent at the mean, 17 percent at the 25^th percentile, and 32 percent at the 10^th percentile. They find that higher debt partially offsets the savings increase, but automatic enrollment still increases net wealth.

Goda and others (2018) examine the effect of a change in the default investment fund of the Thrift Savings Plan (TSP), the federal employees' defined contribution plan, on saving behavior. In September 2015, the TSP default for new hires switched from a low-risk, low-return government securities fund to a lifecycle fund that automatically adjusts its ratio of stock and bond investments based on the proximity of the employee's projected retirement date. The authors find that employees with the lifecycle-fund default are more likely to passively accept both that fund and the default contribution rate than were those with the old default. Interestingly, the default contribution rate that more employees passively accept under the new policy is lower, on average, than the contributions that employees formerly made with an active choice. So although the change resulted in more people defaulting into the lifecycle fund, it also reduced the amount being contributed to the plan by some employees.

Brown, Farrell, and Weisbenner (2016) look at employees who are offered a defined benefit plan as a default, but may choose a defined contribution plan instead. The authors find that individuals with higher income and higher net worth were substantially less likely to choose the default, as were women, those with higher self-assessed investment skills, and those with greater knowledge of the retirement system. When asked if they could go back in time and remake their original pension choice, individuals who were defaulted into the defined benefit plan are about 20 percentage points more likely to regret the choice. Regret is even higher for those prone to procrastination. The findings from this study and Goda and others (2018) point to the need to select default parameters carefully.

To help people who are already enrolled in retirement savings plans to overcome potentially adverse behavioral influences, Beshears and others (2014b) explore whether individuals might increase their retirement saving if they have the option to defer the increase to a later date. The authors find that employees given a delay option exhibit lower savings rates over the ensuing months. However, when framing the delay as being linked to a psychologically meaningful moment, such as the employee's next birthday, the negative effect of offering the delay option is undone. Chalmers and others (2017) examine the effects of message framing and delivery on people's decisions about retirement savings. They find that potential savers are more responsive to messages about retirement saving delivered by members of the same sex, and that descriptive messages (“what others do”) are more effective than injunctive messages (“what one should do”).

Chetty and others (2014) analyze how tax subsidies, employer-provided pensions, and saving mandates affect different types of savers in Denmark. The Danish retirement system is broadly similar to that of the United States, consisting of individual accounts, employer-provided pensions, and a government social security program. The authors estimate that 85 percent of the individuals studied are passive savers, for whom tax subsidies do little to increase saving because action is required to take advantage of them. Automatic contributions, however, are effective, because no action is necessary, yet savings still increase. Active savers are more likely to offset either policy through transfers of assets across accounts, but without increasing net saving. The authors estimate that each $1 of tax expenditure on subsidies increases total saving by 1 cent. By contrast, policies that raise retirement contributions automatically, even if individuals take no action, increase wealth accumulation substantially.

Beshears and others (2014a; 2017) examine illiquidity—that is, restriction on the ability to spend down balances—in a retirement saving system. Because human behavior is subject to the temptation to overspend in the moment, a “socially optimal” retirement system balances illiquidity (to protect assets for the future) against flexibility (for unexpected financial needs now). The authors find that an optimal retirement-saving system is well approximated by just two accounts, one being liquid before retirement and the other being illiquid before retirement.

Beshears and others (2018) analyze what federal employees do with their accumulated TSP savings when they leave their jobs. The authors estimate that more than one-third of those people roll their TSP balances over into individual retirement accounts (IRAs), which very likely have higher fees than the TSP. People with more education are slightly more likely to roll their savings over to an IRA but are also slightly more likely to choose lower-fee IRAs when they do. Goda, Jones, and Ramnath (2016) examine how people respond when they reach age 59½ and thereby become exempt from the 10-percent withdrawal penalty on IRAs. The authors find that crossing that age threshold leads to a $1,500 increase in average annual IRA distributions.

Continuing a stream of past research on the behavioral determinants of saving, this collection of studies highlights the significant effect of “paths of least resistance” on financial decision-making, whether in terms of automatic enrollment, default contribution rates, default portfolio allocations, or other plan provisions. As with choosing when to claim Social Security benefits, the framing of information about saving options and implications influences behavior.

Discussion

Social Security was enacted in 1935 and has served since then as the foundation of retirement income in the United States. The financial challenges facing the program in the coming years are substantial. The youngest cohorts of the baby boom generation will soon transition into OASI eligibility, driving a sizable share of OASI's current and near-term enrollment growth. The Census Bureau (2017, Table 5) projects that the U.S. population aged 62 or older will rise by nearly 25 percent in the next 10 years while the adult population younger than 62 will rise by only 3 percent. Further, the Social Security trust funds' costs are projected to exceed income for the first time in 2020 (Board of Trustees 2019). Under the trustees' intermediate assumptions, OASI trust fund reserves will decline thereafter and, unless Congress enacts reforms, the reserves will be depleted in 2034. Both the importance of the program to people's well-being and the financial challenges it faces going forward motivate the research activities of the NBER RRC and its continuing participation in the Retirement and Disability Research Consortium.

The research of the NBER RRC from 2013 to 2018 offers several overarching takeaways. First, Social Security policy and the way it is communicated to covered workers strongly influences work, retirement, and claiming behavior. Second, the relationships between socioeconomic status, health, and financial well-being in retirement are strong and persistent, making them important considerations in informing Social Security policy. And third, the extent to which people prepare for their financial needs in later life, beyond Social Security, is strongly influenced by the psychological paths of least resistance that powerfully influence saving and other decisions over the life course.