Chapter 13 retirement decision; current influences on the timing of retirement among older workers

CHAPTER 13 Retirement Decision: Current Influences on the Timing of Retirement among Older Workers* Gaobo Pang, Mark J Warshawsky, and Ben Weitzer CONTENTS 13.1 I ntroduction 13.2 L iterature Review 13.3 D ata 13.4 Empirical Results: Estimating and Explaining the Probability of Retirement 13.4.1 D emographics 13.4.2 Retirement Plan Coverage 13.4.3 W ealth Adequacy 13.4.4 Earnings Prospect or Opportunity Cost 288 291 293 296 301 302 302 303 * Opinions expressed here are the authors’ own and not necessarily those of their affil iations We thank Wendi Bukowitz, Carl Hess, Richard Jackson, Allen Jacobson, Erika Kummernuss, Michael Orszag, Ja mes Poterba, Mark Ru loff, Ken Steiner, a nd seminar participants at t he Center for Strategic and International Studies for useful comments 287 © 2010 by Taylor and Francis Group, LLC 288 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling 13.4.5 Social Security Rules 13.4.6 Implication of Business Cycle for DC Plan Participants 305 13.4.7 H ealth Insurance 13.4.8 Rob ustness Tests 13.5 C onclusions Appendix References 304 306 308 309 10 13 T his ch a pter i nv esti gat es t he i nfluences on re tirement b ehavior among older workers who were surveyed by the Health and Retirement Study (1992–2004) It i s f ound t hat i ncreases i n a ll c ategories o f w ealth (pension, housing equity, and other financial wealth) raise the probability of retiring, while good earning prospects induce continued employment Retirement plan types have significant impacts: workers covered by defined benefit (DB) plans are more likely to retire, while the defined contribution (DC) plan coverage delays retirement The probability and thus the timing of retirement for DC plan participants are susceptible to the influence of business cycles through retirement plan income flow fluctuations that are due to i nvestment per formance a nd i nterest rate cha nges Health i nsurance (HI), if conditional on employment, strongly defers retirement, while alternative sources of insurance, such a s employer-sponsored retiree HI, spouse’s H I, public H I, or COBRA coverage, encourage labor force ex it The increases in the full retirement age for social security act to encourage younger cohorts to work longer Keywords: Retirement, pens ion i nvestment, defined benefit, defined c ontribution, s ocial s ecurity, he alth i nsurance, bu siness cycle JEL Classifications: J26, H55, J32, E32, H51 13.1 INTRODUCTION The fac tors a ffecting work ers’ re tirement b ehavior h ave at tracted m uch attention among academia and policy makers This issue deserves renewed research attention and deeper understanding, given recent developments such as the decline in defined benefit (DB) pension plans and the shift to defined contribution (DC) plans, ongoing social security (SS) reforms, and exploding health care costs for retirees as well as for workers © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 289 This chapter investigates the determinants of retirement behavior among older w orkers t hat w ere su rveyed b y t he H ealth a nd Re tirement S tudy (HRS, 1992–2004) Our analysis includes both conventional explanatory variables and new variables to reflect recent environmental changes We revisit issues deemed important in previous studies and add new insights to t he r etirement l iterature F irst, o ur d ata f ollows t he em ployment– retirement behavior of older workers for up to a dozen years Second, our modeling of the ongoing SS retirement age changes reveals the significant policy-driven r etirement differences ac ross co horts Third, we c omprehensively model all major sources of health insurance (HI) coverage and identify their varying impacts This approach may avert the omitted variable bias that could otherwise occur from examining individual factors in isolation Fourth, besides the finding of a significant difference in retirement timing between DB and DC p lan participants, our construction of the DC wealth-earnings replacement rate provides a unique way to gauge the susceptibility of DC plan participants to stock market and interest rate fluctuations It is found that increases in all categories of wealth accumulation (retirement plan, housing equity, and other financial wealth) increase the probability o f r etiring, b ut d ifferentially, wh ile g ood e arning p rospects, implying high opportunity cost for retirement, induce continued employment I t i s w orth n oting t hat o ur co nstruction o f e arning p rospects o r opportunity cost s f orms a n a lternative b ut a m ore st raightforward a nd significant way to incorporate the forward-looking incentives for continued employment (such as the DB or SS benefit accrual), which are shown by some studies to be important Retirement plan types have significant impacts on retirement: besides the nearly universal SS, workers who are entitled to DB plan benefits are more likely to retire t han t hose who a re not, while t he DC p lan coverage delays retirement This phenomenon is presumably in part because many DB plans have work d isincentives beyond certain ages while DC plans are largely age neutral, and in part because most DB plans provide a more secure retirement income flow, lowering overall household exposure to risk There is a concern that the retirement behavior of the DC plan–covered workers is sensitive to stock market boom and bust Our analysis incorporates business cycle effects, as part of the total DC plan effect, by including in come flow fluctuations t hat a re d ue t o i nvestment per formance and ma rket i nterest r ate c ycles These a re r isks pa rticular t o DC p lan © 2010 by Taylor and Francis Group, LLC 290 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling participants We find new evidence to support the above hypothesis; that is, the probability and thus the timing of retirement for DC plan participants are susceptible to the influence of business cycles Workers who have significant income loss (gain) in their DC plans are less (more) likely to retire This may impose some challenges to employers in workforce management When there are market booms, DC plan participants retire just when companies need to add w orkers and when t here are market busts, DC plan participants stay at work just when companies want to cut the workforce Regarding t he i mpact of H I coverage on t he retirement decision, our study reveals that HI, if conditional on employment, strongly discourages retirement, wh ile a lternative sources of H I, such a s employer-sponsored retiree H I (RHI), spouse’s H I, or public H I, fac ilitate or may encourage labor force exit This finding highlights the importance for employers, in pursuit of strategies for human resource management, to consider retirement i ncentives i nherent i n pension plans jointly w ith t he benefits provided by other programs such as HI It should also be noted that benefit modifications f or r etirees (such a s en hancing/eliminating re tiree he alth care coverage) may alter the retirement incentives for current employees Various studies have investigated t he i mportance of SS benefits as an explanation for early retirement In this respect, we have rigorously incorporated the cohort-specific actuarial adjustment factors of SS benefits as defined by the law Our analysis finds that the retirement behavior is significantly linked to such public policies The ongoing increase in the normal retirement age for SS will encourage younger cohorts to work longer We c arefully i ncorporate va rious dem ographic cha racteristics i n t he regressions to control for the heterogeneity of retirement behavior, to the extent a llowed b y t he d ata We h owever ack nowledge t hat o ur r educed form model may bear some insufficiency in addressing the probably simultaneous determinations of savings and labor supply, joint retirement decisions of couples, job mobility, and the availability of pension and health care coverage, and other endogeneities Structural models have been used by researchers to deal with the endogeneity issue and have advanced the understanding of re tirement b ehavior i n s ome d irections These models, h owever, a re often confined to one particular aspect of behavior or the environment due to their complexity, and furthermore bear the risk of biased parameter speci fications Reduced form models, wh ich c an be more t ransparent a nd comprehensive, ser ve a s u seful complements a nd guides to structural models © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 291 This chapter is organized as follows Section 13.2 reviews the relevant literature, Section 3.3 de scribes t he d ata, Section 3.4 d iscusses t he regression results, and Section 13.5 concludes 13.2 LITERATURE REVIEW There is a la rge literature exploring the potential determinants of retirement a mong o lder w orkers Ma ny i nsightful t heoretical a nd em pirical research findings ve co ntributed t o a de eper u nderstanding o f t his complex issue Yet, many debates and questions can be a nswered by new methods and data As a recent example, essays in Madrian et al (2007), all using the HRS surveys, examine retirement prospects, health status and HI, as well as wealth and asset investments for baby boomers This section highlights briefly t he m ost po licy-relevant st rands of research a nd d oes not intend to make the review complete One line of research has investigated the importance of wealth accrual and pension coverage on the timing of retirement Stock and Wise (1990) argue that workers have a n incentive to remain in continued employment until certain ages (often the early retirement ages in pension plans) if the expected gain in utility from postponing retirement outweighs the value of immediate retirement C oile a nd Gr uber (2000) ex amine t he SS i ncentives for retirement and argue that it is in workers’ best interest to stay on the job so as to maximize the SS wealth accrual—the “peak value.” Samwick (1998) also finds that the accrual rate of retirement wealth is a significant determinant of the probability of retirement He argues that the rapidly growing pension coverage and SS entitlements since the 1940s could be the underlying cause of the decline in labor force participation in the early postwar period A second direction of research looks at the impact of pension types on retirement behavior Friedberg and Webb (2005) argue that DB plans tend to have age-related work (dis)incentives that first discourage and later encourage retirement, which contribute to early retirement and lead DB-covered workers t o retire a lmost y ears e arlier on average, compared t o workers with DC p lans Munnell et a l (2004) st udy how pensions a ffect expected and actual retirement ages Regarding the actual retirement decision, they find that pension wealth increases the probability of retiring, while opportunities for more pension accruals lower the probability, and that DB coverage per se raises the probability of actual retirement, while DC coverage reduces the probability Based on survey data about faculty retirement expectations, Flaherty (2006) finds that individuals in DC-only plan situations expect to retire nearly a year later than those in a DB plan, in the context that earlier © 2010 by Taylor and Francis Group, LLC 292 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling voluntary en rollments i n D B o r DC p lans b y fac ulty m embers t o so me degree reveal their differential retirement preferences DB plans provide a steady stream of guaranteed income, while DC plans place participants in considerable exposure to investment and longevity risks The decline in DB plans and the shift to DC plans in the past decades have aroused concern that DC plan–covered workers are vulnerable to business cycles or stock market booms and busts The empirical findings thus far are inconclusive, however Cheng and French (2000) estimate that about 15% of individuals aged 55 and over had a n unanticipated wealth increase of $50,000 or more in constant 1999 dollars between year-ends of 1994 and 1999 The labor force pa rticipation rates a mong t hem, however, i ncreased in t his ma rket boo m pe riod The a uthors bel ieve t hat t he r un-up i n t he stock market was not the primary determinant of employment changes in those years They conjecture t hat t he cha nges may well be a ttributable to the improved employment opportunities and wages in the strong economy and the reduction of work disincentives in the SS system Coile and Levine (2006) f ocus o n a ggregate t rends i n labo r su pply r ather t han t he w ealth effects on i ndividual re tirements a nd find no evidence that stock market changes were the driving force in labor supply because, first, few households have substantial stock holdings, and second, they must be extremely responsive to market fluctuations to generate the observed aggregate employment reversal in the recession (i.e., an increase in the labor force participation rate for older workers aged 55–64 between 2000 and 2002) By contrast, some other studies find evidence that stock market fluctuations alter retirement behavior Coronado and Perozek (2003) study the impact of the stock market boom on retirement decisions, explaining the difference be tween ac tual a nd ex pected r etirement a ges They find that HRS respondents who held equity prior to t he bull ma rket of t he 1990s retired, on average, months earlier than other respondents when the market values increased Hermes and Ghilarducci (2006), using current population survey data, show that the 40% decline in the S&P500 since January 2000 caused the labor force participation of older workers aged 55–64 to increase by 2.64% and 5.36% for men and women, respectively Thei r separate estimate on HRS data shows that the probability of retirement for men aged 61–64 with DC plans fell by 10.7 percentage points from 1998 to 2002 However, their results are sensitive to the age ranges selected.* * The magnitude and sign of the interaction term between DC coverage and year 2002 in their probit mo del i s c onditional on t he i ndependent v ariables S ee A i a nd Norton (2003) for a discussion of the econometrics © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 293 Another important strand of research is devoted to the effect of HI on retirement decisions Gustman and Steinmeier (1994), based on the thenmodest employer contribution cost to employee HI ($2500 per year before age 65), find a small effect of employer-provided HI on retirement behavior Rust and Phelan (1997), explicitly modeling individual risk aversion and a distribution of health care expenditures in a dynamic lifecycle framework, find strong impacts of HI and Medicare on retirement, that is, a significant fraction of “HI-constrained” individuals “optimally” remain employed to attain HI coverage until they are eligible for Medicare coverage at age 65 Blau and Gilleskie (2001) show that the availability of employer-provided RHI i ncreases t he r ate of labor ex it Blau a nd Gi lleskie (2008) similarly show that the access and restrictions to R HI and Medicare have a m odest impact on employment behavior French and Jones (2004) argue that the value of employer-provided HI not only lies in cost reduction but also in u ncertainty reduction for employees Their simulations project t hat a rise in Medicare eligibility age will significantly delay retirement, if workers have no other source of insurance but that tied to employment Rust (2005) simulates faculty retirement decisions and shows that an elimination of the retiree health plan (or a substantial reduction of its generosity) as a cost -cutting measure may significantly reduce t he i ncentive for t he existing faculty to retire Mulvey and Nyce (2005) show that, besides DB pension plans, the availability of RHI boosts the likelihood of early retirement a nd t hat l inking t he employer-paid i nsurance premium to ser vice tenure would mitigate such early exits 13.3 DATA In this analysis, we use the longitudinal, cross-section data from the HRS waves 1992–2004 The d ata se t i s r epresentative of t he na tional po pulation of older workers a nd retirees a nd provides detailed i nformation on demographics, health status and insurance coverage, income and wealth, and em ployment o r r etirement st atus abo ut A mericans o ver t he a ge o f 50.* The respondents and their spouses are (re)interviewed every years We ex clude t hose o bservations t hat lack a w ork–retirement t ransition Specifically, t he “AHEAD” cohort respondents ( born i n 1924 or before) and early baby boomers (born in 1948–1953) are dropped because the former were generally already in the retirement phase when surveyed while the latter were added to the HRS survey in 2004 and therefore have only * Source: http://hrsonline.isr.umich.edu The data and documentation are the Rand version G © 2010 by Taylor and Francis Group, LLC 294 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling one o bservation We f ocus o n t he r etirement beha vior o f o lder w orkers in t he p rivate sec tor a nd ex clude g overnment w orkers (determined, f or each worker, by the job with longest tenure) The analysis is focused on the retirement decision for those respondents aged between 50 and 75 The final data set consists of the following cohorts: HRS (born in 1931– 1941) with survey data 1992–2004, Children of Depression (born in 1924– 1930) with survey data 1998–2004, and War Babies (born in 1942–1947) with survey data 1998–2004 Each respondent therefore has up to seven observations An implicit assumption here is that retirement is reversible– survey respondents may return to work (although not necessarily with the same firm) after retirement This is not a st ringent assumption given the documented findings of retirement reversal in the literature Ruhm (1990) and Mae stas (2007) e ach find t hat about a q uarter of retirees moved t o “unretirement,” while many others reversed from full retirement to partial r etirement C han a nd S tevens ( 2008) sh ow t hat abo ut o ne-third o f older individuals who are ever partially or f ully retired in t he HRS data have reversed their retirement status Retirement is the result of a complex decision-making process Various economic a nd demographic factors a re ex pected to jointly influence the timing a nd t he ex tent of t he t ransition f rom employment to retirement, or a reversal when applicable We run probit regressions to identify factors that may help explain the probability of retirement at any age As retirement behavior is person- or hou sehold-specific, demographic characteristics naturally play an important role In the empirical study, we control for such fac tors as age, gender, educational attainment level, marital status, spo use em ployment st atus, g ood o r bad h ealth, sel f-employment, occupation, union membership, longevity expectation, and employment– retirement transition through a bridge job Particularly interesting are the impacts of the following broader environmental factors: retirement plan coverage, vulnerability of DC plan accounts to the business cycle, wealth adequacy, earning prospects, SS rules, and HI coverage The a ppendix de scribes t he co nstruction o f a ll va riables u sed i n t he regression a nalysis Table 3.1 su mmarizes t he ba sic st atistics o f H RS respondents who are included in the regressions These demographic and financial statistics indicate that the data sample is fairly representative of the po pulation o f o lder w orkers The H RS r espondent-level w eights a re used in the regressions All wealth and income values are in constant 2004 dollars and the unit of measurement is $10,000 © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 295 TABLE 13.1 Summary Data Statistics Variable Retired Age Male Married Spouse retired Bad health Probability of living to age 75+ (%) Self-employed Current job physically challenging Bridge job Union member High school degree or GED Some college education College degree and above DB coverage only DC coverage only Both DB and DC coverage Health insurance, conditional on employment Health insurance, unconditional on employment Health insurance, public Health insurance, COBRA SS early retirement benefits reduction (−%) SS wealth Defined benefit pension wealth SS + DB wealth Median Mean Standard Deviation Minimum Maximum 58 0 75 0.23 57.9 0.46 0.75 0.22 0.13 67.8 0.42 4.17 0.50 0.43 0.41 0.34 26.8 50 0 0 74 1 1 100 0 0.13 0.23 0.34 0.42 0 1 0 0.04 0.22 0.38 0.20 0.41 0.49 0 1 0.23 0.42 0.24 0.43 0 0.17 0.30 0.35 0.37 0.46 0.48 0 1 1 0.62 0.48 1 0.58 0.49 0.12 0.33 0.08 0.27 −50 −39.42 17.92 −50 15.72 14.05 16.74 23.82 7.32 34.11 0 66.5 1485.5 21.32 28.66 28.87 1521.1 (continued) © 2010 by Taylor and Francis Group, LLC 296 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling TABLE 13.1 (continued) Summary Data Statistics Variable DC plan wealth Net housing equity Nonhousing financial wealth Total household wealth Earnings prospect or opportunity cost Total household income DC wealth-earnings replacement rate (%) Median Mean Standard Deviation Minimum Maximum 1.50 8.23 1.88 8.00 12.15 10.02 22.84 29.96 42.55 −638.4 −91.9 2163.9 2280.7 3675.5 41.91 58.84 77.43 −617.8 4791.7 3.21 4.17 5.56 404.6 6.03 8.16 10.68 593.0 2.92 16.02 43.58 400.0 Source: Authors’ calculations based on HRS 1992–2004 survey data Note: Wealth and income variables are in constant 2004 terms and the unit is $10,000 13.4 EMPIRICAL RESULTS: ESTIMATING AND EXPLAINING THE PROBABILITY OF RETIREMENT In t he probit regression model, t he binary dependent va riable t akes t he value of or 1, indicating that the survey respondent is “not retired” or “retired,” respectively, if the HRS labor force participation data indicates so We group partial and full retirements together in the main regression, but also treat them separately in an ordered probit model; the results are not sensitive to the alternative specification (see Section 13.4.8) Respondents are a lso classified as retired if t hey are older t han 65 but t he labor force participation status is missing, or if they are not in labor force and older than 62 Observations are dropped if respondents are disabled or if they are not in labor force when younger than 62 The variable value of otherwise indicates full- or part-time employment including those unemployed but looking for a f ull t ime job.* These steps a im to exclude pe ople who have never worked or are only weakly attached to the labor market so as to allow a valid employment–retirement transition The regressors are those relevant variables listed earlier Table 13.2 reports the regression results for * The separation of partial employment from partial retirement is directly based on t he HRS definition: “If he/she is working part-time and mentions retirement,” the labor force status of the survey respondent “is set to partly retired” and “if there is no mention of retirement,” the status “ is set to work ing part-time.” Source: R AND HRS Data Documentation, Version G, March 2007, p 1035 © 2010 by Taylor and Francis Group, LLC TABLE 13.2 (continued) Probit Regression Results—Marginal Effects on Retirement Decision (Unless Noted Otherwise) Independent Variable Married Spouse retired Probability of living to age 75+ Age Age squared Age cubed Age 61 dummy Age 62 dummy Age 63 dummy Age 64 dummy Age 65 dummy Age 66 dummy No of observations Observation P Pred P (at x-bar) Wald chi2 Prob > chi2 Log pseudo likelihood Pseudo R2 dF/dx −0.06393 0.16794 −0.00027 0.03051 — — — — — — — — 29409 0.21574 0.14761 4914.33 0.00 −10863.4 0.2916 z −8.71*** 22.98*** −2.72*** 24.12*** — — — — — — — — Specification Nonlinear in Age dF/dx −0.06254 0.16515 −0.00026 −0.49796 0.01121 −0.00008 — — — — — — 29409 0.21574 0.14415 4826.86 0.00 −10832.3 0.2937 z −8.64*** 22.87*** −2.72*** −1.02 1.35 −1.64 — — — — — — Specification Age Dummies dF/dx −0.06177 0.16527 −0.00025 0.03069 — — −0.00850 0.19105 0.25921 0.34528 0.15018 0.21479 29409 0.21574 0.14577 5025.38 0.00 −10801.4 0.2957 z Specification Ordered Probit dF/dx −8.47*** −0.25030 22.70*** 0.59307 −2.59*** −0.00096 20.14*** 0.13432 — — — — −0.84 — 2.41** — 2.52** — 2.70*** — 1.15 — 1.52 — 29409 — — 5732.86 0.00 −14931 0.2342 z −8.69*** 23.64*** −2.36** 25.17*** — — — — — — — — Source: Authors’ regression results based on HRS 1992–2004 survey data Notes: Wealth and income variables are in constant 2004 terms and the unit is $10,000 The dependent variable in Specifications 1–6 takes value = retired and = not retired It takes value = fully retired, = partially retired, and = not retired in Specification 7, where results shown are coefficients, not marginal effects The HRS respondent-level weights are used in the probit regressions *, **, and *** denote significance at the 0.10, 0.05, and 0.01 level, respectively; insignificant otherwise © 2010 by Taylor and Francis Group, LLC 300 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling Specification Benchmark Retirement Decision ◾ 301 the benchmark and alternative specifications The probit marginal effects on most of the variables are statistically significant at 1% or 5% level and others at 10% The pseudo R2 is roughly 0.3 in all regressions, which suggests t hat o ur eco nometric spec ifications fit t he d ata w ell The average retirement probability observed in the data is about 22% and our models predict a probability of about 15% using average values of all variables We first report briefly t he results on t he demographic control va riables a nd then highlight the key findings in the benchmark specification that may be more policy relevant 13.4.1 Demographics Various demographic variables are found to have an influence on retirement decisions The probability of retirement increases when workers get older (about 3.1 percentage points for year increase in age) Workers who have higher educational attainment are male, have bad health, or are union members, tend to have higher probabilities of retiring at any age Selfemployed workers a nd ma rried workers a re more l ikely to work longer On the other hand, married workers often show some accord in the retirement timing with their spouses—workers have a greater tendency to stop working when their spouses are retired Life expectancy affects retirement decisions—workers anticipating a good chance of living to advanced ages are likely to retire later A so mewhat su rprising finding i s t hat w orkers ten d t o ve a l ower retirement p robability wh en t heir c urrent j ob i s p hysically cha llenging; this is despite the inclusion of health status, wealth and income variables, and other characteristics in the regression Similarly, Munnell et al (2004) also find that this dummy variable increases actual retirement age, though it has a n egative impact on expected retirement age “Self-selection” may be a plausible (but untested) explanation That is, these workers may actually prefer to remain active after years of physical work We also control for the situations where workers have taken a bridge job, given its importance in the employment–retirement transition [see Ruhm (1990), a mong others] The dummy va riable takes va lue i f t he respondent’s tenure on t he c urrent j ob i s no more t han y ears a nd i s planning to retire in years, or is older than 60 when the planned retirement year is missing The dummy is also set to if the worker has retired from a bridge job Not surprisingly, workers who have taken a bridge job have a higher probability of retiring simply because they have already been in the transition process © 2010 by Taylor and Francis Group, LLC 302 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling 13.4.2 Retirement Plan Coverage Recent changes in private pensions, especially the increasing reliance on DC plans for the provision of retiree benefits, have attracted widespread interest regarding whether the shift is shaking up retirement patterns DB plans generally use a set formula to calculate retirement benefits based on salary and number of service years Regardless of capital market conditions, i n a t raditional DB p lan, r etirees r eceive a ste ady flow o f i ncome benefits for as long as they are alive DC plan participants, by contrast, bear t he risks of investment fluctuations a nd u ncertain longevity These risks may cause DC plan participants to become more cautious in their work decisions We identify the retirement plan coverage based on the value of wealth in that plan form, which is attributable to all work histories, not necessarily c onfined t o t he c urrent j ob S pecifically, a perso n i s co vered b y a DB plan i f t he present va lue of h is or her f uture or c urrent DB benefits is greater than zero Similarly, a positive DC account balance means that there is DC coverage In the regression, we find that the DB pension plan coverage s a s ignificant p ositive e ffect o n t he r etirement zard r ate Besides the nearly universal SS coverage, workers who are entitled to DB plan b enefits a re more l ikely t o retire t han t hose who a re not, w ith t he probability of retiring at any age being 4.5 percentage points higher The impact of the DC plan coverage, however, is negative and is also statistically significant Workers, whose non-SS retirement income is mainly DC plan wealth, exhibit a lower probability of labor force exit by 3.0 percentage points The se findings are consistent with Friedberg and Webb (2005) and the regression’s marginal effects are slightly bigger than in Munnell et al (2004, Table 4) 13.4.3 Wealth Adequacy Standard l ife c ycle t heory postulates t hat re tirement i s t he we alth decumulation phase i n contrast to t he wealth c reation phase i n t he working years Wealth adequacy, among other things, determines whether such a transition is desirable and whether a reasonable standard of living can be maintained in the retirement years, which can be ever-increasing in cost with co ntinued r apid i nflation o f h ealth c are a nd l ong-term c are cost s Wealth here is defined a t t he h ousehold l evel, a s r eported i n t he H RS This m easure i s p robably t he m ost r elevant bec ause r etiring i ndividual workers would consider the entire household’s economic situation as the background rather than the wealth held separately under each household © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 303 member’s name Nevertheless, we adjust household wealth for household size in a n a lternative specification The results a re not sensitive to t hese different wealth measures (see Section 13.4.8) We include four categories of wealth in the regression without a priori imposing a common coefficient on them SS and DB plan wealth, both measured by present discounted values of future benefit payouts, are pooled together given their similar annuity nature This combination helps capture the impact of a secure income source regardless of household wealth concentration in a DB plan or SS DC wealth includes both DC plan and individual retirement account (IRA) account balances Net housing equity refers to t he va lue of primary a nd secondary houses less mortgages a nd home loans Nonhousing financial w ealth i ncludes a ll o ther h ousehold financial a ssets n et o f deb t, ex cluding h ousing eq uity, r etirement p lans, and IRAs The regression results show that all types of wealth help encourage retirement, which is by no means surprising Interestingly, they have differential impacts An improvement in wealth adequacy, for instance, a $100,000 windfall in the above four categories, would imply higher probability of retiring by approximately 0.8, 0.4, 0.1, and 0.3 percentage points, respectively This reveals that housing equity has played a smaller role in financing retirement, relative to other resources As new financial products de velop, pa rticularly i n t he r everse m ortgage ma rket, t he h ousing wealth might be expected to have a bigger influence on retirement 13.4.4 Earnings Prospect or Opportunity Cost Individuals are compensated when they are working and their labor earnings form the main source for consumption and savings The transition to a full (partial) retirement implies the cessation (reduction) of such e arnings, which can be viewed as the opportunity cost for retirees The higher the earnings prospect, the higher is the opportunity cost Presumably, the probability of retirement is lower when the marginal gain of working an extra year is large In our construction, the opportunity cost for a switch to a partial retirement is the full-time earnings last year, while the opportunity cost for a complete labor force exit is the previous full- or part-time earnings, depending on whether he or she was working full or part time last year The opportunity cost is adjusted by a growth/depreciation factor estimated on the current population survey (CPS) March surveys (1989–2007) to account for g eneral in creases o r d eclines in e arnings ca pacity w ith a ge, b y g ender The calculation of these factors is immune to cohort differences in earning © 2010 by Taylor and Francis Group, LLC 304 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling levels and is aggregated across cohort age–earning profiles.* For instance, real annual earnings for male workers aged 70 declines by 1% on average The regression shows that an earnings prospect at the level of $100,000, implying s ignificant o pportunity cost f or r etirement, w ould r educe t he probability of retiring by roughly 3.2 percentage points, other things equal This sh ows a t rade-off f or o lder w orkers i n t hat co ntinued employment leads to more earnings and higher pension and wealth accruals while the income effect at the same time increases the desire for leisure (retirement) The concepts of “option value” in Stock and Wise (1990) and “peak value” in Coile and Gruber (2000) similarly argue that it is rational for workers to continue employed until certain ages in order to maximize DB pension or SS wealth accrual Our construction of the “opportunity cost” offers a straightforward a lternative t hat effectively measures t he financial incentives associated with additional work 13.4.5 Social Security Rules The U.S population is aging and fewer workers are projected to contribute to the SS system relative to the number of beneficiaries As one of the measures, arising from the 1983 reforms, already being used to improve the financial status of SS, the full (normal) retirement age (FRA) is being gradually increased from 65 to 67 Various studies have conjectured that this move will induce more employment among older workers To capture t he i mpact of SS r ules on retirement behavior, we i nclude the early retirement benefit reduction in percentage terms as a r egressor (negative values) This variable is simply the permanent actuarial benefit reduction f actor d efined b y soc ial sec urity ad ministration ( SSA) ba sed on birth year, not based on whether or when a person actually claims SS benefits in the HRS surveys For instance, it takes the value of −20% for a person born in 1937 with FRA 65 if SS retirement benefit starts at age 62, −13.3% at age 63, −6.7% at age 64, and 0% at age 65, while a person born in 1938 and a person born in 1960 with SS benefit commencement at age 62 would receive less SS benefits by 20.83% and 30%, respectively, because * The earnings growth/depreciation factors are calculated in several steps: (1) calculate average full-time earnings of workers by age and gender for each cohort (e.g., age-25 workers in CPS1989, age-26 workers in CPS1990, …, and age-44 workers in CPS2007 are considered as one birth cohort); (2) calculate earnings growth rate at each age for each cohort; (3) average growth rates across birth cohorts; (4) use a polynomial of age to fit the observed growth rates; and (5) t hese fitted a nd smo other g rowth r ates a re t he f actors u sed for t he adjustment on opportunity cost © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 305 their FRAs are 65 and months and 67, respectively.* This variable, which is attributable to exogenous public policy and is independent of the endogenous timing of SS claim, helps pin down the cohort effect The regression result suggests that a reduction of 10 percentage points in SS benefit implies a l ower probability of retiring by approximately 1.1 percentage po ints P erhaps t he e asiest wa y t o u nderstand t his effect is look at the possible retirement decisions between the 1937 and 1943 birth cohorts Their FR As a re a nd 6, r espectively Their r espective b enefit r eductions a re 0% and 25% a t t he e arliest r etirement a ge This percentage point difference imposed by the year difference in FRA would imply a lower probability of retiring at age 62 for the 1943 cohort by more than a half percentage point, ceteris paribus The alternative Specification 6, which a llows for t he i rregular retirement spikes over ages 61–66, would suggest a m uch la rger effect: a pproximately per centage po int d ifference in retirement probability between the 1937 and 1943 cohorts That is, younger cohorts will probably exhibit less early retirement.† 13.4.6 Implication of Business Cycle for DC Plan Participants Compared to DB plans, DC plan values are vulnerable to business cycles Although DC pa rticipants ma y ve t he cha nce t o r eceive subst antial investment returns in a booming capital market, their retirement transition ma y t urn o ut t o be q uite b umpy wh en i nvestment per formance i s poor In addition to the possible shrinkage of DC wealth, a DC participant is also faced with market fluctuations of interest rates if he or she were to purchase a life annuity in the commercial market, a necessary product for insurance against longevity risk To r eflect the effects of business cycle and the timing of annuity purchase, we calculate the lifetime payouts that can be generated by the DC wealth a t ma rket a nnuity p rices, ex pressed a s a per centage o f a verage earnings over survey years 1992–2004 In the calculation, a joint and survivor life table is used (full benefit to survivor) Interest rates are the going * This variable is somewhat arbitrarily set to −50% for work ers you nger t han a nd t hus ineligible for S ocial Security 0% is clearly not appl icable Regression results a re not s ensitive to alternative values such as −100% (a stringent value to represent ineligibility) or −40% (a value just next to −30% that is applicable to the 1960 and younger cohorts) † Other changes in social security rules also affect hours worked or retirement reversal among the older workers Gustman a nd Steinmeier (2008), for i nstance, show t hat t he removal of earnings test increased f ull t ime work of 65 –67 year old m arried men by a bout 2% points between 1992 and 2004, and increased their labor force participation by between 1.4% and 2.2% points, depending on age © 2010 by Taylor and Francis Group, LLC 306 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling 30 year treasury bond yields in the survey years A higher replacement rate implies better preparation for retirement To isolate the timing effect, the same duration of annuity payout is used (age 65–100) regardless of the respondent’s age—otherwise this variable would include an age effect This DC plan payout-earnings replacement rate is similar in notion to the SS and DB pension payouts as percentage of preretirement income, which are commonly used to measure financial preparedness for retirement It is worth noting that the actual annuitization is not necessarily required for this measure to validly reflect the DC plan exposure to business cycles Rather broadly, a ny re tirees w ho l ive o ff interest income from their DC account (e.g., an interest-only strategy) will be a ffected by changes in interest rates in the same direction as with annuity factors Controlling for bond yields directly in the regressions, seemingly a more straightforward way to reflect business cycle, would fail to capture the magnifying effect of high annuity prices when they coincide with poor DC investment performances.* This va riable s a st atistically s ignificant effect on re tirement Specifically, a d rop of 10 percentage points i n t he replacement r ate, due to poor investment performance on DC wealth or due to a hike in annuity p urchase p rices i n t he ma rket, i mplies a l ower p robability o f r etirement i n a ny y ear b y a pproximately per centage po ints This finding, with c areful co ntrol f or o ther fac tors, su pports t he co njecture t hat DC plan participants are affected by ma rket booms a nd busts i n t heir i ndividual retirement, which is in line with Coronado and Perozek (2003) and Hermes a nd Gh ilarducci (2006) The ma gnitude o f i mpact, h owever, i s somewhat smaller than theirs Coile and Levine (2006) find more limited impact of business cycles on aggregate labor force participation 13.4.7 Health Insurance For most workers, employer-sponsored HI is an integral part of the benefits package Such HI coverage may be particularly valuable to the older workers bec ause t he p robability o f fa lling i ll a nd t he r isk o f su ffering catastrophic health care costs generally increase with age Underwriting in t he ma rket for i ndividual H I policies a lso looms la rger at t hese a ges Health care costs and insurance premiums have been on an upward trend in recent decades Absent employer sponsorship or another source of support, workers upon retirement may face steep increases in health care and * Technically, this DC plan wealth-earnings replacement rate can be viewed as an interaction term between two regressors (DC wealth and labor earnings) as commonly used in regressions, but in a division form here and adjusted by an annuity factor © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 307 insurance cost s, wh ich i s l ikely t o d iscourage e arly r etirement On t he other hand, if an employer provides RHI, its employees need not be tied to continued employment and thus have more flexibility in retirement timing Alternatively, a worker covered by HI via his or her spouse’s employer also has such flexibility Another critical channel of HI is through the government-run public insurance a nd welfare programs For i nstance, Medicare, a f ederal H I program, covers most people aged 65 or older and some people younger than 65 with disabilities Medicaid covers people with limited income or resources These public programs help support and may actually encourage retirement (beyond a certain age like 65 in the case of Medicare) Under certain circumstances such as employment termination and other events, workers have the right to continue their group HI on a temporary basis, paying up to 102% of the premium, according to the Consolidated Omnibus Budget Reconciliation Act (COBRA) The premium, determined on a group basis, is generally lower, because it is subsidized by other plan members, for the older workers compared with HI on the retail individual market That is, the general COBRA eligibility provides an alternative way, absent R HI or spousal H I coverage, for retiring workers to bridge t heir transition into retirement especially in ages 63–65 prior to Medicare eligibility bec ause CO BRA g enerally st ipulates a ma ximum o f 18 m onths continuation coverage, ended with eligibility for Medicare In t he regressions, we i nclude a d ummy va riable to i ndicate whether a HRS respondent is covered by HI v ia his or her own employer This is employment-based HI coverage.* A separate dummy variable reflects whether t he r espondent s acc ess t o H I co verage t hat i s n ot t ied t o employment, t hat is, he or she is eligible for R HI coverage by his or her previous or current employer, or he or she is covered by t he health plan sponsored by his or her spouse’ employer Another dummy variable is used to indicate whether the respondent is covered by Medicare, Medicaid, VA/ CHAMPUS (a health benefits program managed by the U.S Department of Veterans Affairs), or other government HI It is observed that these types of HI coverage work as substitutes—many workers switch to an HI coverage delinked to employment in working years as well as upon retirement As public HI mainly eases life in retirement, the RHI and spouse’s * The take-up of employer-provided health insurance is to some extent simultaneous with the employment–retirement decision Bre aking t his endogeneity by c onditioning on l ast p eriod’s HI coverage, however, is infeasible because, in the biennial HRS survey, the HI coverage years ago may not accurately indicate this year’s HI eligibility © 2010 by Taylor and Francis Group, LLC 308 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling HI coverage help bridge the transition from employment to early retirement if a worker is not eligible for Medicare yet As an alternative, COBRA also facilitates the transition and a dummy variable is used to indicate this option for workers aged 63–65.* The regression result shows that HI coverage, if conditional on employment, significantly discourages retirement—the probability of retiring in any year being approximately 16.1 percentage points lower When HI coverage is available through own employer RHI or spouse’s HI, the worker is more likely to retire—the probability is roughly 7.4 percentage points higher Government-sponsored HI programs have an unsurprisingly positive effect on retirement—a higher probability of retirement by approximately 9.3 percentage points for those under such coverage The COBRA option boosts retirement by about 2.0 percentage points The r egression a pparently r eveals subst antial i mpact o f a ll ma jor sources o f H I o n r etirement This finding h ighlights t he i mportance o f integrating retirement incentives in pension plans with the benefits provided by HI A plausible further inference from this finding is that, for an employer to s trategically ma nage human resources, enhancing or eliminating health care coverage for its retirees may have significant impact on current employees’ retirement incentives 13.4.8 Robustness Tests To c heck t he r obustness of ou r re gression re sults, we t est v arious s pecifications W e a lternatively de termine pens ion p lan co verage b y w ealth dominance ( Specification , wh ere “ DB(DC) co verage” i ndicates 5+% of retirement wealth bei ng i n DB(DC) plan a nd “ both DB a nd DC co verage” i ndicates so mewhere i n be tween), ad just w ealth f or ma rried co uples to account for “economies of scale” in their joint consumption by dividing household wealth in all categories by square root of (Specification 3), adjust wealth as in Specification and exclude observations on spouses to restrain the r egressions o n o ne perso n i n t he s ame h ousehold ( Specification 4), allow for pos sible nonlinear effect of a ge by i ncluding t he a ge polynomial (Specification 5) or add age dummies to allow for retirement spikes at specific ages (Specification 6), and run an ordered probit model to treat partial and full retirements separately (Specification 7) We also run a random effects probit * The HRS data does not provide information about COBRA take-up among workers nor about whether their employers had 20 or more employees as required by COBRA In a parsimonious way, we assume that all workers aged 63–65 are eligible for COBRA if they currently have or in the previous survey year had employer-provided health insurance © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 309 model on t he pa nel d ata to add itionally u se t he t ime-dimension i nformation (relative to the above pooled probit regressions across sections),* subtract the self-reported employer and employee contributions from the calculation of DC wealth-earnings replacement rate to further isolate the business cycle impact, and use an additional variable to reflect retirement age flexibility in DB plans (defined as the difference between earliest and FRAs, perhaps an indication o f e arly r etirement in centives).† These a lternative spec ifications only show slight changes in coefficients and maintain the same signs.‡ 13.5 CONCLUSIONS This chapter identifies empirically the critical determinants of retirement behavior of older workers after carefully controlling for demographic factors The findings reveal the importance of considering all influences on the retirement decision in that no single factor in isolation can explain the employment–retirement transition The analysis shows that the likelihood o f r etirement i ncreases w ith w ealth adeq uacy a nd dec reases w ith improved employment opportunities and earning prospects More importantly, the findings in this study may suggest that the traditional pattern of labor force participation rates by age is now undergoing significant transition Workers u nder DC p lans ten d t o r etire la ter a nd t heir t iming of retirement is sensitive to business cycles Changes in the SS rules such as a hike of normal retirement age will push for a longer work career for the younger cohorts Exploding health care costs ma ke employer-sponsored HI increasingly valuable, which, in the context of the elimination of RHI by some employers,§ may turn out to strongly induce continued employment until eligibility for Medicare at age 65 Employers, w hen c rafting benefit pack ages t o st rategically ma nage human r esources, a nd p ublic po licy ma kers, wh en de signing po licies t o * A fi xed effects probit model is inappropriate because it fails to distinguish the effects of those time-invariant va riables A r andom effects pro bit mo del, ho wever, m ay y ield i nconsistent estimates because the assumption that the regressors are independent of error terms does not necessarily hold † The testing power of t he last t wo experiments may be weakened by t he limited data availability Th is analysis uses the self-reported DC plan contributions in the public HRS data fi le It is also difficult to directly identify whether DB benefit payouts are actuarially favorable for early retirees ‡ Results not shown here are available from authors upon request § Mulvey and Nyce (2005, p 119) document that the share of medium and large employers who sponsored some retiree health insurance declined from over 80% before 1980 to only 40% by 2003 © 2010 by Taylor and Francis Group, LLC 310 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling improve worker well-being, need to take into account all benefit programs in a n integrated ma nner To retain older but st ill productive workers a nd promote an orderly retirement process, employers and the government may want to carefully tailor their retirement schemes to meet both the economy’s and employees’ needs in a dynamic and competitive labor market APPENDIX: DATA AND VARIABLE DESCRIPTION Variable Name Binary dependent variable Retirement Retirement plan coverage DB only DC only Both DB and DC Wealth ($10,000) SS and DB wealth DC wealth Net housing equity Nonhousing financial wealth Impact of business cycle DC replacement rate Definition This variable equals to if a respondent is partially or fully retired when the labor force participation data indicates so Respondents are also treated as retired if they are older than 65 but the labor force status is missing, or if they are not in labor force and older than 62 Observations are dropped once the respondents are disabled, or if they are not in labor force when younger than 62 The variable is equal to otherwise, including those unemployed but looking for a full time job (Omitted reference: SS coverage only) Dummy variable equal to if the respondent is covered by the DB plan only (based on DB and DC wealth), otherwise Dummy variable equal to if the respondent is covered by the DC plan only (based on DB and DC wealth), otherwise Dummy variable equal to if the respondent is covered by both DB and DC plans (based on DB and DC wealth), otherwise Household wealth in SS and DB pension plan (present discounted value) Household wealth in DC plan and IRAs Value of primary and secondary houses less mortgages and home loans Total household assets—all debt—housing equity, excluding IRA This variable is the annuity payouts that can be generated by the DC wealth at market annuity prices, expressed as a percentage of average earnings over survey years 1992–2004 It reflects the effects of business cycle and the timing of annuity purchase A higher replacement rate implies better preparation for retirement © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 311 Variable Name Definition In the annuity calculation, a joint and survivor life table is used (full benefit to survivor) Interest rates are the going 30 year treasury bond rates in the survey years To isolate the timing effect, the same duration of annuity payout is used (65–100) regardless of the respondent’s age; otherwise this variable would include the age effect SS impact Early retirement benefit reduction (%, non-positive value) Earnings prospect ($10,000) Earnings or opportunity cost Health insurance coverage HI conditional on R’s employment HI unconditional on R’s employment (Partially cohort effect) This is the permanent actuarial benefit reduction factor in percentage if the respondent claims SS earlier than FRA defined by SSA based on the birth year For instance, it takes value −20% if a person born in 1937 starts claming SS benefit at age 62, −13.3% at age 63, −6.7% at age 64, and 0% at age 65, while a person born in 1938 and a person born in 1960 who start claiming SS at age 62 would receive less SS benefits by 20.83% and 30%, respectively, because their FRAs are 65 and months and 67, respectively This variable is set to −50% for respondents younger than 62 because they are not eligible for SS (0% inappropriate) The regression results are not sensitive to alternative values such as −40%, −100%, or others This variable reflects the earnings if the respondent chooses to work or opportunity cost if he or she chooses to retire For a full-time worker, this variable equals earnings (wage/salary income + bonuses/overtime pay/ commissions/tips + 2nd job or military reserve earnings, professional practice or trade income) The opportunity cost for a switch to a partial retirement is the full-time earnings last year, while the opportunity cost for a full retirement is the previous full- or part-time earnings, whichever applicable In the construction of opportunity costs, previous earnings are adjusted by a growth/depreciation factor, which is calculated on CPS March surveys 1989–2007 by gender and age Dummy variable equal to if the respondent is covered by HI via his or her own employer, otherwise Dummy variable equal to if the respondent is covered by RHI via his or her own employer, or by his or her spouse’s HI or RHI, otherwise (continued) © 2010 by Taylor and Francis Group, LLC 312 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling (continued) Variable Name HI via public program COBRA Demographics Age Male Bad health Self-employed Current job physically challenging Bridge job Union member Married Spouse retired High school diploma or GED Some college education College degree and above Longevity expectation Definition Dummy variable equal to if the respondent is covered by Medicare, Medicaid, VA/CHAMPUS, or other government HI, otherwise Dummy variable equal to if the respondent is aged 63–65 and currently have (or had in last survey year) employer provided HI, otherwise Actual age Dummy variable equal to if the respondent is male, otherwise Dummy variable equal to if self-reported health status is fair or poor, otherwise (good, very good, and excellent) Dummy variable equal to if the respondent is self-employed, otherwise Dummy variable equal to if current job requires lots of physical effort, lifting heavy loads, or stooping/ kneeling/crouching), otherwise The value for retirees is determined by their previous job Dummy variable equal to if the respondent’s tenure on current job is no more than years and is planning to retire in years, or is older than 60 when the planned retirement year is missing, otherwise The dummy is set to if a worker has retired from a bridge job defined here Dummy variable equal to if the respondent is covered by labor union, otherwise The value for retirees is determined by their previous job Dummy variable equal to if the respondent is married, otherwise Dummy variable equal to if the respondent’s spouse is retired, otherwise Dummy variable equal to if the respondent has high school or GED degree, otherwise Dummy variable equal to if the respondent has some college education (degree less than a BA), otherwise Dummy variable equal to if the respondent has a BA degree or greater, otherwise The omitted category is education below high school Self-reported subjective probability (0–100%) of living to age 75+ Source: Authors’ constructions based on HRS 1992–2004 survey data Notes: The data source is the HRS waves 1992–2004 (Rand version G) This table describes the construction of all va riables used in t he regressions All llar values are in constant 2004 terms and the unit is $10,000 © 2010 by Taylor and Francis Group, LLC Retirement Decision ◾ 313 REFERENCES Ai, C a nd E C N orton 2003 I nteraction t erms in logi t a nd p robit mo dels, Economics Letters, 80(1), 123–129 Blau, D M and D B Gilleskie 2001 Retiree health insurance and the labor force behavior of older men in the 1990s, Review of Economics and Statistics, 83(1), 64–80 Blau, D M a nd D B Gilleskie 2008 The role of retiree health insurance in t he employment behavior of older men, International Economic Review, 49(2), 475–514 Chan, S a nd A H S tevens 2008 I s r etirement b eing r emade? D evelopments in la bor ma rket pa tterns a t o lder ag es, in Managing Reti rement P ayouts, J Ameriks and O Mitchell (eds.), Oxford University Press, Oxford, U.K Cheng, I.-H and E French 2000 The effect of the run-up in the stock market on labor supply, Economic Perspectives, Issue Quarter IV, 48–65 Coile, C C a nd J Gruber 2000 S ocial security incentives for retirement, NBER Working Paper No 7651 Coile, C C and P B Levine 2006 Bulls, bears, and retirement behavior, Industrial and Labor Relations Review, April 59(3), 408–429 Coronado, J L a nd M P erozek 2003 W ealth effects a nd t he co nsumption o f leisure: Retirement decisions during t he st ock ma rket b oom of t he 1990s, Finance and Economics Discussion Series No 2003–20, Board of Governors of the Federal Reserve System Flaherty, C 2006 The impact of pension plan type on expected retirement age, Working Paper, Stanford University, Stanford, CA French, E and J B Jones 2004 The effects of health insurance and self-insurance on retirement behavior, Working Paper No 2004–12, Center for Retirement Research at Boston College, Boston, MA Friedberg, L and A Webb 2005 Retirement and the evolution of pension structure, Journal of Human Resources, Spring 40(2), 281–308 Gustman, A a nd T Steinmeier 1994 Em ployer-provided he alth in surance and retirement behavior, Industrial and Labor Relations Review, 48(1), 124–140 Gustman, A and T Steinmeier 2008 How changes in social security affect recent retirement trends, NBER Working Paper No 14105 Hermes, S and T Ghilarducci 2006 The effect of the stock market crash on retirement decisio ns, D epartment o f Eco nomics Working pa per, U niversity o f Notre Dame, Notre Dame, IN Madrian, B., O S Mitchell, and B J Soldo (eds.) 2007 Redefining Retirement: How Will Boomers Fare? Oxford University Press, Oxford, U.K Maestas, N 2007 Back to work: Expectations and realizations of work after retirement, Rand Working Paper no WR-196–2 Mulvey, J and S Nyce 2005 Strategies to retain older workers, in Reinventing the Retirement Paradigm, Robert L Cla rk and Olivia S M itchell (eds.), O xford University Press, New York, pp 111–132 Munnell, A H., R K T riest, a nd N A J ivan 2004 H ow p ensions a ffect expected a nd ac tual retirement ag es, Working Paper No 2004–27, C enter for Retirement Research at Boston College, Boston, MA © 2010 by Taylor and Francis Group, LLC 314 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling Ruhm, C J 1990 Bridge jobs and partial retirement, Journal of Labor Economics, 8(4), 482–501 Rust, J 2005 I mpact o f r etiree he alth p lans o n fac ulty r etirement decisio ns, in Recruitment, Re tention a nd Re tirement: The Three R’s o f H igher Ed ucation in t he 21s t Cen tury, Rob ert Cla rk a nd J ennifer M a (eds.), Ed ward E lgar, Northhampton, MA, pp 135–169 Rust, J and C Phela n 1997 How social security and medicare affect retirement behavior in a World of incomplete markets, Econometrica, 65(4), 781–831 Samwick, A 1998 N ew evidence on pensions, social security, and the timing of retirement, Journal of Public Economics, 70(2), 207–236 Stock, J H and D A Wise 1990 Pensions, the option value of work, and retirement, Econometrica, September 58(5), 1151–1180 © 2010 by Taylor and Francis Group, LLC ... reveal the importance of considering all influences on the retirement decision in that no single factor in isolation can explain the employment retirement transition The analysis shows that the. .. maintain the same signs.‡ 13. 5 CONCLUSIONS This chapter identifies empirically the critical determinants of retirement behavior of older workers after carefully controlling for demographic factors The. .. LLC Retirement Decision ◾ 293 Another important strand of research is devoted to the effect of HI on retirement decisions Gustman and Steinmeier (1994), based on the thenmodest employer contribution