Work Opportunities for Older Americans Series 2, February 2006 DOES WORKING LONGER MAKE PEOPLE HEALTHIER AND HAPPIER? By Esteban Calvo* Introduction Findings from Previous Studies Financing retirement is one of the major challenges Much of the literature relevant to older workers and facing an aging U.S. population. If individuals con- well-being is actually focused on other issues, such as tinue to retire in their early 60s, many will be hard healthy aging and retirement and not on the effects of pressed to maintain an adequate standard of living work per se. Studies of healthy aging suggest that, throughout retirement due to the declining role of for many individuals, working to an older age could Social Security, the shift to 401(k) plans, and low per- have a number of positive physical and psychological sonal saving rates. Combine the retirement income effects. The primary reason is that opportunities for crunch with the dramatic increase in life expectancy, active engagement tend to decline in later life, and and continued employment in later life appears to be this trend tends to adversely impact overall well- an attractive option. being. 1 Since most jobs require workers to engage in While it is clear that working longer would benefit a number of productive and social activities, working older Americans financially, less attention has longer might bring a number of benefits associated focused on the non-monetary effects of work at older with these activities. Prolonged employment during ages. This brief addresses the impact of late-life paid later life allows many older workers to continue doing work on physical and psychological well-being. The activities similar to those performed in middle-age. first section reviews the literature on work at older In this sense, working longer appears to be helpful in ages and elderly well-being. The second section maintaining meaning and a sense of purpose in life, describes the analysis. The third and fourth sections as well as adapting to aging.2 present the results. The fifth section identifies vul- Research on retirement, as opposed to work, also nerable groups. A final section offers concluding provides important clues regarding the potential con- thoughts. sequences of working longer. Work is fundamental to one's identity; therefore, leaving one's job implies a partial loss of identity, which may produce lower lev- els of psychological well-being. Conversely, retire- * Esteban Calvo is a graduate student in Sociology at Boston College and a graduate research assistant at the Center for Retirement Research at Boston College. The author thanks the following people for their comments on and other forms of help LEARN MORE in connection with this brief: Paula Errázuriz, Andy Eschtruth, Margaret Ford, Jessica Johnson, Alicia H. Munnell, Steven Search for other publications on this topic at: Sass, Mauricio Soto, and John Williamson. This brief was www.bc.edu/crr/pub_wor.shtml adapted from a longer article by Esteban Calvo (forthcoming). 2 Center for Retirement Research ment can be viewed as a gain in freedom to pursue Methodology other activities or interests, thus enhancing older individuals' well-being. To the extent that retirement The following analysis assesses the impact of individ- implies the loss of valued rewards, it will be viewed uals' work status in 2000 on their physical and psy- negatively and perceived as a difficult adjustment. chological well-being in 2002. It uses longitudinal But retirement is also an experience of release, data drawn from the Health and Retirement Study though its benefits vary between long-term and (HRS) and the RAND-HRS data base.1 0 The sample recent retirees. 3 Another important factor in the indi- analyzed is composed of individuals aged 59 to 69 vidual’s response to retirement is the subjective sense who were working or not-working in the year 2000 of control. Voluntarily leaving a job could provide a and were alive in 2002. People dead in 2002 were sense of control in the work/retirement decision, included in the regressions analyzing mortality. 1 1 thus moderating any negative aspects of retirement. 4 The variable "paid work status" measures if peo- Although retirement may not be a choice for some ple are working for pay or not. It is related to the particular groups (e.g. people constrained by financial variable labor force participation, but is not the same. or health problems), the vast majority of retirees cur- As an illustration, for the sample analyzed, 74 per- rently leave their jobs voluntarily. 5 cent of those not working for pay are retired, 18 per- Although research on healthy aging and retire- cent are not in the labor force (e.g. women who never ment provides valuable insights, only a few studies worked outside of the home), 6 percent are disabled, have directly addressed the effects of late-life work on and 2 percent are unemployed or partially retired. the physical and psychological well-being of older Well-being was assessed by using a comprehen- Americans. Most of these studies have found posi- sive set of six measures:12 tive effects of work on well-being. The workplace appears as an important source of friends, confidants, . self-rated health; and social support, even after long-term retirement as . self-rated memory; retirees tend to retain the network of friends they . activities of daily living (ADL) such as developed when they were working.6 Job-related bathing, eating, and dressing; rewards seem to have a positive impact on mental . instrumental activities of daily living (IADL) health.7 Performing more than 100 annual hours of such as handling money, shopping, and volunteer or paid work beyond the retirement age has cooking; an independent beneficial effect on older adults' self- . mood indicators such as depression, loneli- rated health and survival.8 It has also been suggested ness, and happiness; and that the conditions of work are more influential on . mortality. 13 physical and mental health than the pure amount of work. Namely, older employees who work in low- The study controls for previous well-being by stress jobs with the hours they desire, experience bet- including status in 1998.14 It also includes variables ter health. And surprisingly, physically demanding for healthy behaviors, such as whether the respon- jobs have a positive effect on the physical health of dent smokes, the number of drinks the respondent those aged 65 and over. 9 has per day, and whether the respondent exercises Most of the studies mentioned above identified vigorously. Finally, this study also controls for demo- relationships between work and well-being using data graphic and socioeconomic factors. 15 from a single point in time. These cross-sectional In sum, the methodology of this study enables us studies cannot establish causal relationships, because to determine whether health problems cause cessa- people with higher well-being may be more likely to tion of work or continued work causes better health remain at work. For example, individuals with no (see Figure 1). physical disability are more likely to continue work- ing than those with a physical limitation. Only longi- How Work at Older Ages Affects tudinal studies — which follow the same individuals over time — can differentiate whether health prob- Health lems cause cessation of work or continued work caus- es better health. The analysis below builds on the The results of the analysis of the impact of later-life previous studies, but uses a longitudinal design, work on well-being are presented in Appendix Table 1 includes more control variables, and expands the (Model 1 through Model 3). They show that those notion of well-being to include more indicators. who worked in 2000 tended to report greater well- being in 2002 than those who did not work in 2000. 3 Center for Retirement Research significantly affect self-rated memory, illustrating that F IGURE 1. TIME PERIODS U SED IN DATA ANALYSIS other factors are better predictors than work (e.g. 1998 2000 2002 being male and initial self-rated memory). Not surprisingly, well-being baselines and healthy Baseline behaviors are related to follow-up physical and psy- Work Status Health chological well-being. But after controlling for base- Health line effects and other variables, these findings sug- Demographic gest that there is a causal relationship between work Factors and well-being. 17 Socioeconomic How Undesirable Jobs Affect Factors Psychological Health Source: Author’s methodology. The results presented above suggest that longer work- For example, Figure 2 illustrates that paid work at ing lives have beneficial effects on individuals' physi- older ages reduced the probability of reporting cal and psychological well-being. However, this fair/poor health in 2002 by 6 percent. Other vari- analysis did not distinguish between different types ables that had a statistically significant impact on the of jobs. Previous literature suggests that the charac- probability of reporting fair/poor health are high- teristics of work matter, and we all know that jobs are lighted using dark bars. While working, exercising, not all the same. Work can be backbreaking, stress- being white, married, and having more education ful, and tedious. Therefore, this next exercise and income had a favorable impact, poor physical explores the hypothesis that undesirable jobs could and psychological health in 1998, smoking, being have detrimental effects on individuals' well-being. male and Hispanic had an unfavorable impact on "Undesirable jobs" are those that have excessive self-rated health. demands or otherwise cause dissatisfaction. 18 They As illustrated in Appendix Figures 1 through 4, are the type of jobs that people would probably not work also had a modest, but statistically significant, choose voluntarily. 19 beneficial impact in the cases of ADL, IADL, mood The analysis of a subgroup of older workers sug- indicators, and mortality. Paid work at older ages gests that having an undesirable job does not change reduced the probability of having any ADL limitation the favorable effects of paid work on self-rated health, by 2 percent, of having any IADL limitation by 2 per- ADL and IADL (see results for Model 4 in Appendix cent, of reporting negative mood indicators by 2 per- Table 1). However, it does have an impact on follow- cent, and of dying by 1 percent. 16 Paid work did not up mood indicators and mortality. Higher job satis- F IGURE 2. EFFECTS OF PAID WORK AT OLDER AGES ON THE PROBABILITY OF R EPORTING F AIR/P OOR SELF -RELATED HEALTH (2002) Statistically significant Statistically insignificant Work (2000) Bad health (1998) 38% ADL limit (1998) IADL limit (1998) Neg. mood (1998) Bad memory (1998) Diseases (1998) Drinks (2000) Exercise (2000) Smoking (2000) Age (2000) Male (2000) White (2000) Hispanic (2000) Married (2000) Education (2000) Income (2000) -25 -20 -15 -10 -5 0 5 10 15 20 25 Percent Source: Author’s calculations (probit estimation reporting discrete and marginal coefficients) using data from the 1998, 2000, and 2002 waves of the Health and Retirement Study . 4 Center for Retirement Research faction is associated with an improved mood. Those who do not enjoy going to work are 5 percent more F IGURE 3. V ULNERABLE GROUPS BY PAID W ORK S TATUS likely to report negative mood indicators. Lower job demands are statistically associated with survival but have only a trivial effect.2 0 25% 40% Working in desirable job Not working (other than Vulnerable Groups forced) While working longer seems beneficial for most, poli- Not working and forced into Vulnerable retirement cymakers should consider the potential negative con- 20% Groups sequences of forcing all people to work, for example, Working in undesirable job by raising the early eligibility age for Social Security 15% benefits.21 In rough numbers, people who retire involuntarily constitute about 20 percent of the popu- Source: Author’s calculations using data from the 2000 lation analyzed (see Figure 3). For this group, work is wave of the Health and Retirement Study. not an alternative.22 Also, 15 percent of the popula- tion analyzed was working in undesirable jobs, and *Note: The sample size is 6,511. 40 percent was working in less demanding and more satisfying jobs. However, if the early retirement age were raised, an important group of people would be forced to work in late-life, regardless of whether or Conclusion not they enjoy working or have a highly demanding job. For this group, continuing work would have This analysis provides several interesting insights opposing effects: beneficial in terms of self-rated into the relationship between work at older ages and health, ADL and IADL, but harmful for their mood, well-being. It suggests that longer working lives will and to a lesser extent, their survival. For both vulner- help most people maintain their overall well-being. able groups, working fewer annual hours, reducing While working longer seems beneficial for most work-strain, or participating in other social or produc- people, it will likely have negative consequences for tive activities — like volunteering — seem more ben- some. The type of job seems to be a critical factor. eficial than working longer. Undesirable jobs can wash out the potential favorable effects of work. Another critical factor is the opportu- nity to continue working. Older workers may be will- ing to prolong paid work, but, in order to find a job, they need to be able to work and have a real demand for their labor. Policymakers need to consider these factors when evaluating proposals to keep people in the labor force. These findings suggest interesting areas for future research. For example, an increase in the early retirement age reduces the ability of people to volun- tarily decide their labor force participation. Less con- trol in the work/retirement decision could have an adverse impact on the well-being of older individuals.23 Also variations in the benefits of work as people move through their 60s could be examined. Even though work at older ages seems beneficial for many, the benefits may decrease, or stop increasing, after a certain age or amount of time worked per year. 5 Center for Retirement Research Endnotes 1 Much evidence suggests that more active elderly 9 Herzog et al. (1991). people are less likely to die or have various physical and mental illnesses (Brummett et. al., 2001; Everard 10 The HRS is a nationally representative panel sur- et. al., 2000; Rowe and Kahn, 1998; and Siegrist, et vey of U.S. households with individuals aged 51-61 in al., 2004). Remaining active and socially engaged 1992 and their spouses. The RAND Center for the seems to have a positive impact on identity (Erickson Study of Aging created the RAND-HRS data based on et al., 1986). Social integration, social support, and HRS data. positive interactions tend to be associated with posi- tive well-being, even after controlling for other vari- 11 The sample size varies for each dependent variable ables related to social competence (Cohen, 2004). according to available data. The data used in all of the Social and productive activities appear as effective as regressions were adjusted to the weights of the gener- fitness activities in lowering the risk of death (Glass, al population. et. al, 1999). And the risk of social isolation is com- parable to that associated with cigarette smoking 12 The effect of work on these six measures of well- (House, 2001). being (given the set of control variables) was estimat- ed using probit regression models for each well-being 2 Different authors suggest this interpretation, but outcome. Probit regressions fit maximum-likelihood none of them identify a compulsory reason for mal- models, allowing the estimation of the probability of adjustment in retirement (Atchley, 1999; and Rowe each well-being outcome, conditional on a set of inde- and Kahn, 1998). Numerous studies suggest that pendent variables. Rather than reporting the regular non-paid activities, such as volunteering, care-giving, probit coefficients, this brief reports the discrete and housework, and informal help to friends, can be as marginal effects, holding all other variables at their beneficial as paid work for older individuals' well- mean, a technique described in Long and Freese, being. However, they do not provide the benefits of 2004. additional income that paid work does. For research regarding non-paid activities, see: Herzog et al. 13 The six measures of well-being were coded as (1991); Herzog et al. (1998); Glass et al. (1999); Luoh dummy variables. For more information about relat- and Herzog (2002); and Morrow-Howell et al. (2003). ed measurements in the HRS, see Saint Clair et al. (2004); Steffick (2000); Fonda and Herzog (2004); 3 Bossé et al. (1990); Elder (1995); Kim and Moen and Ofstedal et al. (2005). (2002); and Kosloski et al. (2001). 14 Baselines also consider chronic diseases, a binary 4 Volunteerism is an expression of self-efficacy, mas- variable that measures the presence or absence of any tery, self-directedness, self-agency, autonomy, self- of eight chronic diseases: high blood pressure, dia- determination, and other forms of sense of control betes, cancer, lung disease, heart disease, stroke, psy- that are paralleled by biological responses that buffer chiatric problems, and arthritis. More information stress reactions. Similarly, the degree of voluntari- about the measures of chronic diseases can be found ness in the work/retirement decision partially medi- in Saint Clair et al. (2004); and Fisher et al. (2005). ates the well-being outcomes of either work or retire- ment (Bender and Jivan, 2005; Herzog et al., 1998; 15 Gender, race, ethnicity, and marital status were Gallo et al., 2000; and Kim and Moen, 2002). coded as dummy variables. Household income was measured in increments of $100,000 dollars for a 5 According to the HRS (2002), about 30 percent of better graphical presentation of the data. the retirees were forced to retire. Prudential (2005) finds that 38 percent fall into this group. 16 Control variables generally behaved as expected. Generally, the corresponding baseline of the depend- 6 Bossé et al. (1990). ent variable had the higher coefficient. However, self- rated health and smoking were two of the most con- 7 Erickson et al. (1986). stant predictors of well-being outcomes. 8 Additional hours do not increase the benefit. See Luoh and Herzog (2002). 6 Center for Retirement Research 17 Since groups of workers and non-workers were not randomly assigned and data were collected every two years, these results should be interpreted with cau- tion no matter how rigorous the controls included in the mode appear. 18 Job satisfaction measures how much people enjoy going to work. "Work demands" is an index of job characteristics including: physical effort, lifting heavy loads, stooping/kneeling/crouching, good eyesight, and stress. The index runs from 0 to 5. 19 Additional data regarding voluntariness and sense of control in relation to work status and work pat- terns are insufficient for the sample analyzed. 20 These effects are both similar and contrasting to cross-sectional effects reported in Herzog et al. (1998). Job satisfaction — understood as the most intrinsic non-economic job reward — positively affects psychological well-being. However, this study found detrimental effects of physical demands on older adults' mood indicators. These results must be interpreted with special caution, because people will tend to remain at work if they enjoy working and if they do not have a highly demanding job. 21 Raising the early eligibility age could also reduce the flexibility to work part time as a transitional path from full-time work to retirement. 22 Previous studies also suggest that overall well- being is likely to be impaired for those forced into retirement (Bender and Jivan, 2005; Herzog et al., 1998; Gallo et al., 2000; and Kim and Moen, 2002). People are forced into retirement mainly due to health reasons and job displacement (Prudential, 2005). 23 Fortunately, increasing labor force participation and maintaining choice are not incompatible (Herzog et al., 1991). 7 Center for Retirement Research References Atchley, Robert C. 1999. Continuity and Adaptation in Fonda, Stephanie and Regula Herzog. 2004. Aging: Creating Positive Experiences. Baltimore, MD: “Documentation of Physical Functioning Measured Johns Hopkins University Press. in the Health and Retirement Study and Health Dynamics among the Oldest Old.” HRS/AHEAD Bender, Keith A. and Natalia A. Jivan. 2005. What Documentation Report, Ann Arbor: Survey Research Makes Retirees Happy? Issue in Brief 28. Chestnut Hill, Center University of Michigan. MA: Center for Retirement Research at Boston College. Gallo, William T., Elizabeth H. Bradley, Michele Siegel, and Stanislav V. Kasl. 2000. "Health Effects of Bossé, Raymond, Carolyn M. Aldwin, Michael R. Involuntary Job Loss among Older Workers: Findings Levenson, Kathryn Workman-Daniels, and David J. from the Health and Retirement Survey." Journal of Ekerdt. 1990. "Differences in Social Support among Gerontology: Social Sciences 55B(3): S131-S140. Retirees and Workers: Findings from the Normative Aging Study." Psychology and Aging 5(1): 41-47. Glass, Thomas A., Carlos Mendes de Leon, Richard A Marottoli, and Lisa F Berkman. 1999. "Population Brummett, Beverly H., John C. Barefoot, Ilene C. Based Study of Social and Productive Activities as Siegler, Nancy E. Clapp-Channing, Barbara L. Lytle, Predictors of Survival among Elderly Americans." Hayden B Bosworth, Redford B. Williams, and Daniel British Medical Journal (319): 478-483. B. Marx. 2001. "Characteristics of Socially Isolated Patients with Coronary Artery Disease Who Are at Herzog, Regula A., James S. House, and James N. Elevated Risk of Mortality." Psychosomatic Medicine Morgan. 1991. "Relation of Work and Retirement to 63:267-272. Health and Well-being in Older Age." Psychology and Aging 6(2): 201-211. Cohen, Sheldon. 2004. "Social Relationship and Health." American Psychologist 59:676-684. Herzog, Regula A., Hazel R. Markus, Melissa M. Franks, and Diane Holmberg. 1998. "Activities and Elder, G. H., Jr. 1995. “The Life Course Paradigm: Well-being in Older Age: Effects of Self-concept and Social Change and Individual Development.” In Educational Attainment." Psychology and Aging 13(2): Moen, P.; Elder, G. Jr. & Lusher, K. (Eds.), Examining 179-185. Lives in Context: Perspectives on the Ecology of Human Development. Washington, DC: American House, James S. 2001. "Social Isolation Kills, but Psychological Association. How and Why?" Psychosomatic Medicine 63:273-274. Erickson, Erik H., Joan M. Erikson, Helen Q. Kim, Jungmeen E. and Phyllis Moen. 2002. Kivnick. 1986. Vital Involvement in Old Age. New “Retirement Transitions, Gender, and Psychological York: Norton. Well-being: a Life-course, Ecological Model." Journal of Gerontology: Psychological Sciences 57B(3): P212- Everard, Kelly M., Helen W. Lach, Edwin B. Fisher, P222. and Carolyn M. Baum. 2000. "Relationship of Activity and Social Supports to the Functional Health Kosloski, Karl, David Ekerdt, and Stanley De Viney. of Older Adults." Journal of Gerontology: Social Sciences 2001. "The Role of Job-related Rewards in Retirement 55(4): 208-S212. Planning." Journal of Gerontology: Psychological Sciences 56B(3): 160-169. Fisher, Gwenith G., Jessica D. Faul, David R. Weir, and Robert B. Wallace. 2005. “Documentation of Long, J. Scott and Jeremy Freese. 2004. Regression Chronic Diseases Measures in the Health and Models for Categorical Dependent Variables Using Retirement Study (HRS/AHEAD).” HRS/AHEAD STATA. Texas: STATA Press. Documentation Report, Ann Arbor: Survey Research Center University of Michigan. Luoh, Ming-Ching and Regula A Herzog. 2002. "Individual Consequences of Volunteer and Paid Work in Old Age: Health and Mortality." Journal of Health and Social Behavior 43(4): 459-509. 8 Center for Retirement Research Morrow-Howell, Nancy, Jim Hinterlong, Philip Rozario, and Fengyan Tang. 2003. "Effects of Volunteering on the Well-being of Older Adults." Journal of Gerontology: Social Sciences 58B(3): S137- S145. Ofstedal, Mary B., Gwenith G. Fisher., and Regula A. Herzog. 2005. Documentation of Cognitive Functioning Measures in the Health and Retirement Study. Ann Arbor, MI: Survey Research, University of Michigan. Prudential Financial. 2005. Roadblocks to Retirement. A Report on What Happens When Living Today Gets in the Way of Financial Security Tomorrow. Prudential's Four Pillars of Retirement Series. Rowe, John W. and Robert L. Kahn. 1998. Successful Aging. New York: Pantheon. Saint Clair, Patricia, Delia Bugliari, Sandi Chien, Steven Haider, Orla Hayden, Michael Hurd, Serhii Ilchuk, Gabriela Lopez, David Loughran, Constantijn Panis, Philip Pantoja, Monika Reti, Julie Zissimopolous. 2004. RAND HRS Data Documentation, Version D. Labor and Population Program, RAND Center for the Study of Aging. Siegrist, Johannes, Olaf Von Dem Knesebeck, and Craig Evan Pollack. 2004. "Social Productivity and Well-being of Older People: a Sociological Exploration." Social Theory and Health 2:1-17. Steffick, Diane E. 2000. Documentation of Affective Functioning Measures in the Health and Retirement Study. HRS/AHEAD Documentation Report, Ann Arbor: Survey Research Center University of Michigan. University of Michigan. 1998, 2000 and 2002. Health and Retirement Study. http://hrsonline.isr.umich.edu/ Appendix APPENDIX TABLE 1. EFFECTS OF PAID W ORK AT OLDER AGES ON INDIVIDUAL’S F OLLOW-UP WELL -B EING Self-Rated Health (2002) ADL (2002) IADL (2002) Variables (Fair-Poor vs. Excellent-Good) (Any Limitation vs. No Limitation) (Any Limitation vs. No Limitation) Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Working for Pay (2000) -.158*** -.062*** -.058*** - -.094*** -.026*** -.024** - -.080*** -.026*** -.021** - Enjoy Working (2000) - - - -.032 - - - -.004 - - - -.020 Demanding Job (2000) - - - .004 - - - -.002 - - - .000 Fair/Poor Self-Rated Health (1998) .403*** .376*** .322*** .074*** .070*** .069*** .059*** .057*** .016 Any ADL Limitation (1998) .104*** .096*** .064* .255*** .252*** .191*** .082*** .080*** .085*** Any IADL Limitation (1998) .045* .037 -.045 .061*** .062*** .021 .135*** .135*** .118*** Negative Mood Indicators (1998) .095*** .082*** .079*** .037*** .033*** .026* .033*** .032*** .033** Fair/Poor Self-Rated Memory (1998) .084*** .058*** .049** .017* .015 .023* .019** .019** .014 Any Chronic Diseases (1998) .103*** .105*** .059*** .051*** .051*** .034*** .027*** .026** .006 3 Or More Drinks Per Day (1998) -.002 -.017 -.012 -.019 -.020 -.012 -.019 -.020* -.008 Vigorous Exercise (1998) -.040*** -.038*** -.006 -.024*** -.023** -.005 -.016* -.014* -.009 Smoking (1998) .070*** .056*** .038* .018* .017 .016 .025*** .023** .013 Age (2000) -.001 -.003 .001 .001 .001 .000 Male (2000) .042*** .025 .001 .010 .006 .017* White (2000) -.060*** -.050* -.003 .003 -.005 -.009 Hispanic (2000) .056* .019 .011 .005 -.016 -.020 Married (2000) -.026* .003 -.013 .010 -.009 -.005 Years of Education (2000) -.011*** -.010*** -.001 .000 -.001 .000 Household Income (2000) -.020* -.013 -.001 .001 -.015* -.006 Pseudo R 2 .0329 .2931 .3089 .2444 .0320 .2710 .2725 .1984 .0358 .2349 .2398 .1652 Number of Observations 6251 6251 6251 2563 6251 6251 6251 2563 6251 6251 6251 2563 Source: Author’s calculations (probit estimation reporting discrete and marginal effects) using data from the 1998, 2000 and 2002 waves of the Health and Retirement Study. *Note: *p<.05, **p<.01, ***p<.001 APPENDIX TABLE 1. EFFECTS OF PAID W ORK AT OLDER AGES ON INDIVIDUAL’S F OLLOW-UP WELL -B EING (CONTINUATION ) Mood Indicators (2002) Mortality (2002) Self-Rated Memory (2002) Variables (Negative vs. Positive) (Dead vs. Alive) (Fair-Poor vs. Excellent-Good) Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Working for Pay (2000) -.075*** -.026** -.018* - -.026*** -.014*** -.009** - -.090*** -.032** -.023 - Enjoy Working (2000) - - - -.049** - - - -.003 - - - -.013 Demanding Job (2000) - - - .004 - - - .002* - - - .004 Fair/Poor Self-Rated Health (1998) .065*** .055*** .032* .038*** .004*** .014** .124*** .091*** .010*** Any ADL Limitation (1998) .083*** .077*** .033 -.013** .012** .003 .038 .032 .025 Any IADL Limitation (1998) -.021 -.023 -.035 .015* .014* -.003 .038 .029 .006 Negative Mood Indicators (1998) .265*** .243*** .190*** .003 .003 -.002 .060*** .054** .058* Fair/Poor Self-Rated Memory (1998) .045*** .038*** .023 -.006 -.006 -.002 .413*** .382*** .360*** Any Chronic Diseases (1998) .027* .025* .028* .015** .013** .006* .026 .028 .021 3 Or More Drinks Per Day (1998) .001 -.018 -.020 .008 .001 .002 .040 .016 .000 Vigorous Exercise (1998) -.023** -.019* -.018 -.010** -.009** -.003 -.019 -.021 -.032* Smoking (1998) .048*** .037*** .027 .023*** .020*** .017*** .020 .007 -.026 Age (2000) .000 .001 .002*** .001** .000 .001 Male (2000) -.038*** -.019 .016*** .010*** .067*** .075*** White (2000) .002 .005 -.004 .001 -.048** -.032 Hispanic (2000) -.016 .015 -.004 -.005 .019 .014 Married (2000) -.030*** -.046*** -.012** -.012*** .008 -.001 Years of Education (2000) -.007*** -.006** .000 .000 -.021*** -.019*** Household Income (2000) .000 .001 -.007 -.004 -.017 -.031* Pseudo R 2 .0152 .1859 .1996 .1665 .0187 .0909 .1182 .1952 .0094 .1935 .2178 .2101 Number of Observations 6060 6060 6060 6060 6473 6473 6473 2608 6053 6053 6053 2513 Source: Author’s calculations (probit estimation reporting discrete and marginal effects) using data from the 1998, 2000 and 2002 waves of the Health and Retirement Study. *Note: *p<.05, **p<.01, ***p<.001. APPENDIX FIGURE 1. E FFECTS OF PAID W ORK AT OLDER AGES ON THE PROBABILITY OF R EPORTING ANY ADL LIMITATION (2002) Statistically significant Statistically insignificant Work (2000) Bad health (1998) ADL limit (1998) IADL limit (1998) Neg. mood (1998) Bad memory (1998) Diseases (1998) Drinks (2000) Exercise (2000) Smoking (2000) Age (2000) Male (2000) White (2000) Hispanic (2000) Married (2000) Education (2000) Income (2000) -25 -20 -15 -10 -5 0 5 10 15 20 25 Percent Source: Author’s calculations (probit estimation reporting discrete and marginal coefficients) using data from the 1998, 2000, and 2002 waves of the Health and Retirement Study. APPENDIX FIGURE 2. EFFECTS OF PAID W ORK AT OLDER AGES ON THE PROBABILITY OF R EPORTING ANY IADL LIMITATION (2002) Statistically significant Statistically insignificant Work (2000) Bad health (1998) ADL limit (1998) IADL limit (1998) Neg. mood (1998) Bad memory (1998) Diseases (1998) Drinks (2000) Exercise (2000) Smoking (2000) Age (2000) Male (2000) White (2000) Hispanic (2000) Married (2000) Education (2000) Income (2000) -25 -20 -15 -10 -5 0 5 10 15 20 25 Percent Source: Author’s calculations (probit estimation reporting discrete and marginal coefficients) using data from the 1998, 2000, and 2002 waves of the Health and Retirement Study. APPENDIX FIGURE 3. EFFECTS OF PAID W ORK AT OLDER AGES ON THE PROBABILITY OF REPORTING N EGATIVE MOOD INDICATORS (2002) Statistically significant Statistically insignificant Work (2000) Bad health (1998) ADL limit (1998) IADL limit (1998) Neg. mood (1998) Bad memory (1998) Diseases (1998) Drinks (2000) Exercise (2000) Smoking (2000) Age (2000) Male (2000) White (2000) Hispanic (2000) Married (2000) Education (2000) Income (2000) -25 -20 -15 -10 -5 0 5 10 15 20 25 Percent Source: Author’s calculations (probit estimation reporting discrete and marginal coefficients) using data from the 1998, 2000, and 2002 waves of the Health and Retirement Study. APPENDIX FIGURE 4. E FFECTS OF PAID WORK AT OLDER AGES ON THE PROBABILITY OF DYING (2002) Statistically significant Statistically insignificant Work (2000) Bad health (1998) ADL limit (1998) IADL limit (1998) Neg. mood (1998) Bad memory (1998) Diseases (1998) Drinks (2000) Exercise (2000) Smoking (2000) Age (2000) Male (2000) White (2000) Hispanic (2000) Married (2000) Education (2000) Income (2000) -25 -20 -15 -10 -5 0 5 10 15 20 25 Percent Source: Author’s calculations (probit estimation reporting discrete and marginal coefficients) using data from the 1998, 2000, and 2002 waves of the Health and Retirement Study. Issue in Brief 9 About the Center Affiliated Institutions The Center for Retirement Research at Boston American Enterprise Institute College was established in 1998 through a grant from The Brookings Institution the Social Security Administration. The Center's mis- Center for Strategic and International Studies sion is to produce first-class research and forge a Massachusetts Institute of Technology strong link between the academic community and Syracuse University decisionmakers in the public and private sectors Urban Institute around an issue of critical importance to the nation's future. To achieve this mission, the Center sponsors a Contact Information wide variety of research projects, transmits new find- Center for Retirement Research ings to a broad audience, trains new scholars, and Boston College broadens access to valuable data sources. Since its Fulton Hall 550 inception, the Center has established a reputation as Chestnut Hill, MA 02467-3808 an authoritative source of information on all major Phone: (617) 552-1762 aspects of the retirement income debate. Fax: (617) 552-0191 E-mail: crr@bc.edu Website: http://www.bc.edu/crr www.bc.edu/crr/ © 2005, by Trustees of Boston College, Center for The research reported herein was supported by The Atlantic Retirement Research. All rights reserved. Short sections of Philanthropies. The opinions and conclusions expressed are text, not to exceed two paragraphs, may be quoted without solely those of the author and should not be construed as explicit permission provided that the author is identified representing the opinions or policy of The Atlantic and full credit, including copyright notice, is given to Philanthropies or the Center for Retirement Research at Trustees of Boston College, Center for Retirement Research. Boston College.