1 TRAJECTORIES TO RETIREMENT: THE ROLE OF PERSONAL TRAITS, ATTITUDES AND EXPECTATIONS Péter Hudomiet, Andrew Parker and Susann Rohwedder RAND SIEPR Conference on Working Longer and Retirement October 8, 2015 Funded by the Alfred P. Sloan Foundation Widespread “non-traditional” retirement 2 Labor histories of the 56-57 year old full time workers in HRS, followed for 14 years: Full time -> retirement Gradual retirement Unretirement Full time -> part time work Always full time work Unemployment -> retirement Disability -> retirement Total N 972 360 435 358 324 47 91 2587 Percent 37.6 13.9 16.8 13.8 12.5 1.8 3.5 100.0 Large literature on timing of retirement 3 Role of incentives (Social Security, private pensions, health insurance) Role of health Krueger and Pischke (1992), Imbens, Rubin, and Sacerdote (2001), Sevak (2001), Coronado and Perozek (2003), Brown et al. (2010) Role of job characteristics Blau and Gilleskie (2001), McGarry (2004), French (2005), Burkhauser and Cawley (2006), Bound et al. (2010) Role of wealth Gustmant and Stenimeier (1986, 1994, 2000, 2004), Stock and Wise (1990), Krueger and Pischke (1992), Gruber and Wise (1999, 2004), Blau and Gilleskie (2001, 2006, 2008), Nyce et al. (2013), Hurd and McGarry (1993), Blekesaune and Solem (2005), Angrisani et al. (2013) Relatively little on psychological factors Barnes-Farrell (2003), Angrisani et al. (2013), McGonagle et al. (2015) Literature on the role of psychological factors in economic behavior 4 Cognition on economic outcomes Schooling, job success, welfare status, child neglect, poverty, crimes, savings, etc. Jensen (1998), Banks and Oldfield (2007); McArdle, Smith and Willis (2009), Christelis, Jappelli and Padula (2010); Grinblatt, Keloharju and Linnainmaa (2011) Personality on economic outcomes Mortality, divorce, occupations, smoking, health, wealth, earnings, etc. Roberts et al. (2007), Smith (2006), Hampson et al. (2007), Hurd et al. (2012) How might psychological factors impact retirement trajectories? 5 Preference examples • Being with family • Being productive • Being around others • Traveling • Various hobbies Life shock example • Some take better care of themselves • Better health, longer retirement How might psychological factors impact retirement trajectories? 6 Affecting economic opportunities Some might want to…(preferences) keep career job reduce hours on career job find a bridge-job Who can do that? people maybe relevant in career jobs? people with good abilities? with the right personalities? maybe relevant in finding bridge-jobs? or holding on to a career job? In summary: Psychological factors may affect retirement by … 7 Making certain workers WANT to work longer Making employers want to KEEP certain workers Making employers want to HIRE certain workers into bridge jobs Why study effect of psychological factors on retirement trajectories? 8 Pre-determined variables Cognitive abilities & personality relatively stable characteristics Rank-preserving decline in fluid intelligence from early adulthood Hard to influence by policy If psychological factors important predictors, still of interest to policy as it may illuminate: Who likely responds to incentives or other policy levers Who would benefit from what type of additional support In this paper 9 11 waves from the Health and Retirement Study Individuals working full-time at age 56, follow for 14 years Construct realized retirement trajectories Relate realized and expected retirement trajectories to cognitive ability and personality Investigate mechanisms through which psychological factors shape retirement trajectories Most closely related papers 10 Maestas (2010) “Unretirement” is wide-spread and largely expected Many go through non-traditional retirement trajectories. Angrisani et al. (2013) Analyze 2-year labor market transitions Impact of personality traits and job characteristics McGonagle et al. (2015) Analyze perceived work ability and some labor force outcomes Effect of detailed psychological and work factors HRS DATA: Analytic sample 11 56-57 year old between 1992-1998 Full time workers at baseline Follow them for 14 years until age 70-71 Only use values from these 14-year windows Age of interviewees by cohort and wave: Cohort 1935-36 1937-38 1939-40 1941-42 1992 1994 1996 56-57 58-59 60-61 56-57 58-59 56-57 1998 62-63 60-61 58-59 56-57 HRS wave 2000 2002 2004 64-65 66-67 68-69 62-63 64-65 66-67 60-61 62-63 64-65 58-59 60-61 62-63 2006 70-71 68-69 66-67 64-65 2008 2010 2012 70-71 68-69 70-71 66-67 68-69 70-71 Constructing Retirement Trajectories 12 Step 1: Assign labor market status for each person-year observations Step 2: Fill in missings using job start, job end, and retirement dates Step 3: Translate into retirement trajectories Status Working Not working Code Works full-time F Works part-time Retired Unemployed Disabled Dead Other not in labor force P R U D E O Missing . Examples of retirement trajectories: 1. FFFFRRRR 2. FFPP.RRR 3. F...RRPP 4. FPPRRRRE 5. FPPPPPPPF 6. FF..FFFFF Realized retirement trajectories 13 Based on info between age 56 and 70 Retirement trajectory 1 2 3 4 5 6 7 8 9 10 Full time work to retirement Gradual retirement Unretirement Full-time to part-time work Always full time Retired after unemployment Retired after disability Deceased by age 70 Left the sample Other uncategorized cases Examples FFFPPPRR FFFRRRPP FFPURRRR FFFDRRRR FFRRRREE 12.3% 12.6% 3.2% Less than 40% move from full-time to retirement, About 40% move to part-time 14 Distribution of retirement trajectories, N=2,586 Full time work to retirement Gradual retirement Unretirement Full-time to part-time work Always full time work Retired after unemployment Retired after disability % 37.7 13.9 16.8 13.8 12.5 1.8 3.5 Total 100.0 Trends in Retirement Trajectories 15 Add timing of retirement (early, middle, late) Look for trends across cohorts Find trends towards working longer: Most notable increases in fraction … - Working full-time past age 65, then retire - Working full-time, then move to part-time Control for cohort effects Little change in the fraction of “non-traditional” retirement Effect of Psychological Factors 16 Cognitive abilities and Big 5 personality traits Most widely used psychological factors We intent to expand the list of measures Measure of Cognitive Ability 17 27 point score of working and episodic memory Closely linked to fluid intelligence (Del Missier et al., 2013) Sum of immediate and delayed word recall, serial 7 subtraction and backward counting Methodology Age-adjusted person specific mean values, using responses between age 50 and 61 Then we standardize Big 5 Personality Traits 18 Conscientiousness Neuroticism needing attention and social interaction Agreeableness needing intellectual stimulation, change, and variety Extraversion experiencing the world as threatening and beyond control Openness to experience willingness to comply with conventional rules, norms, and standards needing pleasant and harmonious relations with others Methodology 26 questions asked every second wave since 2006 Use person-specific means (typically 2 values) Then we standardize Identification of Effects of Interest 19 Psychological factors predetermined and timeinvariant Assume that psychological factors exogenous to all other controls Regression of retirement trajectories on … psychological factors only will show their total effect psychological factors plus other controls will show relevant mechanisms and remaining effect of psych factors Outline of regression analysis 20 𝒚𝒊 = 𝜷𝟎 +𝜷𝒑 𝒑𝒊 + 𝜷𝒙 𝒙𝒊 + 𝒖𝒊 𝑝𝑖 vector of psychological factors 𝛽𝑝 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑥𝑖 set of controls Demographics (gender, race, education) Health indicators (self-reported at ages 56 and 66; subjective probability of living to age 75) Labor market variables at the main job at age 56 (occupations; DB pensions, DC pensions, health insurance) Marital status (being single at ages 56 and 66) Wealth (log total household wealth) OLS regressions of working past age 65: High cognition, extraversion predict working longer Agreeableness promote retiring earlier 21 [1] 0.046 -0.002 0.068 -0.058 0.009 0.020 [2] 0.042 0.003 0.066 -0.041 0.011 0.005 Cognitive ability Neuroticism Extraversion Agreeableness Conscientiousness Openness to experience Cohort dummies Demographics Health variables Labor market variables Marital status Household wealth R squared 0.030 0.038 Explained cognition 7.9% Explained extraversion 1.6% Explained agreeableness 30.4% N=2,376; bold = stat. significant at 5% [3] 0.036 0.014 0.057 -0.035 0.005 0.003 [4] 0.034 0.018 0.053 -0.033 0.015 -0.005 [5] 0.035 0.018 0.054 -0.033 0.015 -0.007 0.050 20.7% 16.1% 40.0% 0.094 24.6% 22.2% 43.9% 0.095 23.1% 20.1% 43.2% [6] 0.037 0.018 0.054 -0.035 0.014 -0.004 0.098 18.5% 19.9% 39.8% OLS: Working full-time past age 65 Cognition systematic predictor (+), agreeableness in [1] Extraversion no longer significant 22 [1] [2] 0.029 0.035 -0.008 -0.004 0.025 0.022 -0.042 -0.020 0.002 0.004 [3] 0.032 0.001 0.018 -0.018 0.002 [4] 0.032 0.006 0.015 -0.016 0.010 [5] 0.032 0.005 0.015 -0.016 0.010 [6] 0.034 0.005 0.015 -0.020 0.011 0.015 0.011 0.003 0.003 0.005 0.032 0.038 0.088 0.088 0.091 Cognitive ability Neuroticism Extraversion Agreeableness Conscientiousness Openness to 0.026 experience Cohort dummies Demographics Health variables Labor market variables Marital status Household wealth R squared 0.017 N=2,376; bold = stat. significant at 5% Multinomial logit of retirement trajectories, left out category: full retirement 23 Cognitive ability Neuroticism Extraversion Agreeableness Conscientiousness Openness to experience Cohort dummies Demographics Health variables Labor market variables Marital status Household wealth Gradual retire 0.030 -0.050 0.080 -0.122 0.029 -0.029 N=2,477; bold = stat. significant at 5% Unretire 0.125 0.130 0.231 -0.221 -0.023 0.071 Always Full to full-time part-time 0.308 0.221 0.004 0.063 0.263 0.320 -0.112 -0.193 0.086 0.051 0.015 -0.073 Other 0.080 0.090 0.110 0.096 -0.129 0.131 Retirement Expectations 24 Look at several measures. Overall strongly predictive of realized retirement trajectories, but some discrepancies. Retirement Expectations 25 Example: Probability of working full-time past age 65 (P65), measured at baseline Do psychological factors explain retirement expectations? OLS of Expectations: work full-time past age 65 Cognition and openness significant and positive 26 [1] 0.029 -0.004 -0.015 -0.018 -0.018 Cognitive ability Neuroticism Extraversion Agreeableness Conscientiousness Openness to 0.054 experience Cohort dummies Demographics Health variables Labor market variables Marital status Household wealth R squared 0.025 [2] 0.019 -0.004 -0.013 -0.009 -0.017 [3] [4] [5] [6] 0.018 0.020 0.022 0.025 -0.001 0.003 0.004 0.004 -0.016 -0.021 -0.016 -0.015 -0.008 -0.002 -0.003 -0.008 -0.016 -0.007 -0.007 -0.004 0.044 0.038 0.028 0.025 0.028 0.038 0.058 0.147 0.163 0.174 N=2,325; bold = stat. significant at 5% Psychological Factors predict Retirement Expectations and Realizations 27 High cognitive ability predicts working longer in both expectations and realizations Extraversion predicts working longer (part-time) in realizations, not expectations Agreeableness predicts retiring earlier in realizations, not expectations. Openness to Experience predicts working longer in expectation, not realization Implications 28 Psychological factors are hard to influence by policy Because predetermined But they reveal important channels look at mechanisms May inform where policies are most effective Retirement attitudes can be influenced Information campaigns Incentives Part-time jobs an integral part of working longer Extraverts are successful in getting them How can we help introverts? Do they even want part-time jobs? Do they want different types of jobs? Can job matching/job exchanges help? What are the mechanisms through which psych. factors affect retirement trajectories? 29 So far, considering … Retirement attitudes Job characteristics Health Psychological factors strongly predict all three. Thank you! 30 Trends in trajectories: Some evidence for working longer 31 Cohort that turned 56 or 57 year old in year… Full retirement, before age 62 Full retirement, age 62-65 Full retirement, age 66+ Gradual retirement, before age 65 Gradual retirement, age 66+ Unretirement, 1 wave in R Unretirement, 2+ waves in R Always full time work Full-time to part-time work Retirement after unemployment Retirement after disability Total N 1992 8.0 13.0 8.3 2.6 5.9 7.5 4.3 8.4 8.7 1.2 2.6 100.0 667 1994 6.7 12.2 6.9 3.9 6.7 8.5 2.6 10.0 9.7 0.9 2.7 100.0 617 1996 6.4 11.9 8.4 4.9 6.1 8.7 4.1 7.8 10.3 2.1 2.4 100.0 632 1998 5.2 10.4 10.6 3.6 6.5 8.6 4.2 9.9 11.2 1.1 2.4 100.0 675 Subjective probabilities line up with realized retirement trajectories 32 Probability of working full time past age 65 Full retirement, before age 62 12.1% Full retirement, age 62-65 16.1% Full retirement, age 66+ 33.5% Gradual retirement, bef. age 65 13.2% Gradual retirement, age 66+ 30.7% Unretirement, 1 wave in R 29.8% Unretirement, 2+ waves in R 14.0% Always full time work 50.5% Full to part time work 39.7% Retirement after unemployment 34.0% Retirement after disability 23.8% 29.0% Total Mechanism 1: Retirement attitudes Bad things Good things 33 Cognitive Open to Neurotic Extravert Agreeable Conscienti. ability experience One can be own boss One can take it easy One can travel No pressure Time with spouse Time with children One can do hobbies Do voluntary work Not productive There is sickness No income Boring Miss co-workers Inflation + + + + + + + + + + + + + + + + + + Mechanism 1: Retirement attitudes 34 Cognitive Open to Neurotic Extravert Agreeable Conscient. ability experience + Felt forced to R Satisfied with R Reasons for retirement: health issues do other things be with family + + + + + + Mechanism 2: Health 35 Fair/Poor health at Fair/Poor health at Subj. probability age 56 age 66 live to 75 Cognition Neuroticism Extraversion Agreeableness Conscientiousness Openness to experience Cohort dummies Demographics R squared -0.018 0.039 -0.035 0.027 -0.015 -0.031 0.056 -0.056 0.031 -0.051 0.013 -0.024 0.019 -0.009 0.004 0.005 0.019 0.027 0.087 0.126 0.095 N = 2,376; bold = stat. significant at 5% Mechanism 3: Job characteristics 36 Distribution of work hours in jobs after unretirement: • 73 % are part time jobs Overall (in the entire sample) • 40% of people spent time in part-time jobs 1-19 hours 20 hours 21-34 hours 35-40 hours 41+ hours Total N 179 51 81 67 50 427 % 42.0 11.9 18.9 15.6 11.7 100.0 Mechanism 3: Job characteristics 37 Distribution of occupations among those who take part-time jobs • N = 1,142 • Large inflow into service sector (and sales, transportation) • Large outflow from management, production Management and support Professional occupations Sales Office Services Mechanical and production Transport, material moving Other Total % Baseline full-time job New part-time job 18.8 12.5 20.0 18.7 12.2 14.1 13.7 14.0 10.4 16.9 12.1 7.5 7.4 9.4 5.4 7.0 100.0 100.0 Mechanism 3: Job characteristics 38 We look at # of job changes (assuming no return after job-loss) • Reducing hours is hard in career jobs • Many people experience more than one job change Full retirement Gradual retirement Unretirement Always full time work Full-time to part-time work Retirement after unemployment Retirement after disability Average N=2,564 0 82.8 37.4 0.0 54.8 37.8 Number of job changes 1 2+ missing 11.0 5.2 1.1 41.4 20.1 1.1 58.2 35.9 5.9 20.1 14.5 10.6 22.1 31.0 9.1 Total 100.0 100.0 100.0 100.0 100.0 50.9 88.1 52.4 23.6 9.5 26.0 100.0 100.0 100.0 25.5 1.8 17.4 0.0 0.7 4.2