1. Brooke Helppie McFall, Amanda Sonnega and Robert J. Willis University of Michigan Retirement Research Consortium Conference August 6-7, 2015 1

2.  Increased interest in learning about:  Motivation to work longer or retire earlier, including job characteristics  Bridge job employment  What can we learn from more detailed occupational information? 2

3.  To what extent are occupational composition and occupational characteristics related to trends in retirement timing?  Describe changes in occupational composition over time ▪ Within cohorts ▪ Between cohorts  Unpack role of occupation and occupational characteristics in retirement trends 3

4.  Health and Retirement Study (HRS) Core Surveys, 1992-2012  RAND HRS Data Version N  Restricted HRS Occupation Data  Aggregated from ~980 categories to 192, consistent over time, larger cell counts allow sharing of more results  In most years, over 98% of workers have occupation codes (lower in 1994 and 1996)  Linked to Occupational Information Network job characteristics (O*NET)  Use Cobb-Douglas weighting to create measure including both level and importance. Use CPS frequencies to weight components of aggregated occupations 4

5.  Results from HRS detailed occupation data and linked characteristics  Examine:  changes in occupational composition 1992-2012  relationship between occupation and retirement behavior  relationship between job characteristics and retirement behavior 5

6. 6

7. 7

8. 8 Table 1. Top 10 early departure and long working occupations, all cohorts Total these 20 occupations, 1992-20127,140 Total for 191 occupations with min 5, 1992-201228,624 Early departure occupations Change '92- '94 to 2010-12 Change '92-'94 to 2006-08 Occupation titleObs Other managers2,549-36%-34% Shipping and receiving clerks181-42%-47% Other mechanics and repairers244-46%-54% Precision metal working occupations234-82%-75% Farm occupations, except managerial241-43%-26% Other machine operators, assorted materials576-64%-51% Production inspectors, testers, samplers, and weighers298-44%-50% Construction equipment operators161-65%-60% Other freight, stock, and material handlers370-40%-36% Other personal service occupations303-39%-46% Long working occupations Managers of medicine and health occupations194242%168% Other financial specialists252181%103% Management analysts149282%256% Lawyers and Judges141184%133% Health technologists and technicians157129%164% Cust. service reps, investigators and adjust., except insurance 216294%158% Teacher assistants254133%121% Farm operators and managers124128%193% Gardeners and groundskeepers276126%101% Taxi cab drivers and chauffeurs220291%374%

9. Summary statistics for job characteristic variables and analyses 9 SourceVariableDescriptionMeanMedianSt. Dev.Obs. HRSEarly retirementLast occ observed before age 63? (Yes=1, No=0)0.3800.493781 HRSLate retirement“Last” occ observed at age 66+? (Yes=1, No=0)0.4500.53781 HRSMore difficult (jdiff)Job requires doing more difficult things than before (1-4, strongly agree to strongly disagree) 2.5930.83456 HRSLots of stress (jstres)Job involves a lot of stress (1-4, strongly agree to strongly disagree) 2.4530.823658 HRSPhysical effort (jphys)Job requires physical effort (1-4, all/almost all the time to none/almost none) 2.8131.123642 HRSCould reduce hours (credh)Could reduce hours if wanted to (Yes=1, No=0)0.3600.483295 O*NetActivity 4Analyzing data or information (0 - 1)0.480.460.153780 O*NetActivity 5Making decisions and solving problems (0 - 1)0.620.610.133780 O*NetActivity 9Controlling machines and processes (0 - 1)0.380.330.193780 O*NetActivity 11Interacting with computers (0 - 1)0.480.510.213780 O*NetActivity 13Repairing and maintaining electronic equipment0.220.190.123780 O*NetActivity 14Documenting/recording information (0 - 1)0.53 0.153780 O*NetActivity 16Assisting and caring for others (0 - 1)0.460.420.153780 O*NetActivity 17Performing for or working directly with the public0.490.50.183780 O*NetActivity 18Coaching and developing others (0 - 1)0.470.440.153780 O*NetAbility 3Mathematical reasoning (0 - 1)0.34 0.133780 O*NetAbility 4Arm-hand steadiness (0 - 1)0.350.380.173780

10. Which occupations predict “late” retirement? 10 Variablecoef.se Variablecoef.se Financial Managers (excl. cat.)-- General office clerks0.25*0.15 Managers of properties and real estate0.26*0.15 Teacher assistants0.25*0.14 Management analysts0.34*0.18 Industrial machinery repairers-0.10.16 Purchasing managers, agents and buyers; business and promotion agents -0.240.16 Production supervisors or foremen-0.29**0.14 Postsecondary teachers0.25**0.13 Precision metal working occupations-0.130.15 Primary school teachers-0.140.12 Other precision work, assorted materials -0.37**0.15 Social workers0.27*0.16 Farm operators and managers0.37**0.15 Clergy and religious workers0.36**0.15 Textile sewing machine operators-0.140.15 Lawyers and Judges0.52**0.24 Other machine ops, assorted materials-0.32**0.13 Writers, authors, technical writers0.40**0.2 Bus drivers0.25*0.14 Designers0.40**0.2 Taxi cab drivers and chauffeurs0.42***0.14 Musician or composer0.40**0.2 Other freight, stock, & material handlers -0.10.13 Athletes, sports instructors, officials and announcers 0.38*0.21 Guards, watchmen, doorkeepers0.22*0.12 Licensed practical nurses-0.250.17 Other protective services0.41**0.19 Real estate sales occupations0.32**0.16Constant0.48***0.1 Other sales and sales related0.20*0.11R-squared0.15 Messengers0.45***0.17Adjusted R-squared0.09 Observations2842 Linear probability models (OLS) with "late" retirement (0/1) as dependent variables and occupation indicators as regressors. All data from 2010. Includes respondents who were 51-61, working full-time, and not self-employed at their baseline interview, and over age 66 in 2010. "Late" retirement equals 1 if the last recorded occupation was at age 66 or later, or if the respondent was over 66 and still had a listed occupation in 2010. Excluded occupation is "Financial Managers." Only occupations which were statistically significant in one of the two regressions are included. Significance levels denoted as * for p<0.1, ** for p<0.05, *** for p<0.01.

11. Which job characteristics predict “late” retirement? 11 Covariate source: HRS onlyO*Net onlyBoth Variablecoefsecoefsecoefse Physical effort (jphys) 0.04***0.010.04***0.01 Lots of stress (jstres) 0.09***0.010.09***0.01 More difficult (jdiff) 0.06***0.010.05***0.01 Could reduce hours (credh) 0.2***0.020.17***0.02 Analyzing data or information 0.29*0.170.270.18 Making decisions and solving problems 0.070.16-0.190.16 Controlling machines and processes -0.46***0.1-0.33***0.1 Interacting with computers -0.33***0.08-0.22**0.09 Repairing and maintaining electronic equipment -0.030.110.020.12 Documenting/recording information -0.110.12-0.020.12 Assisting and caring for others 0.050.090.17*0.1 Performing for or working directly with the public 0.39***0.050.3***0.06 Coaching and developing others -0.18*0.09-0.040.1 Mathematical reasoning -0.080.11-0.170.11 Arm-hand steadiness 0.170.120.090.12 Constant -0.15***0.040.48***0.07-0.070.08 R-squared 0.110.050.15 Adjusted R-squared 0.110.050.14 Observations 305137803051 Regressions are linear probability models. Dependent variable is "late" retirement indicator. Same sample restrictions as in regressions of retirement on occupation, except these use characteristics of last occupation observed, max one observation per respondent, for those who are observed past age 66. "Late" retirement is equal to 1 if last observed occupation in HRS data was at age 66 or later, and zero otherwise. Significance levels denoted as * for p<0.1, ** for p<0.05, *** for p<0.01.

12. 12  Some interesting compositional change in detailed occupational information  Some occupations are associated with later work  Role of job characteristics

13.  Use hazard model to include time-varying factors in prediction of retirement timing and probability of full-time work past age 65.  Examine relationship between retirement and individual O*Net variables  Case studies of occupations associated with working longer  Which are bridge jobs, which are late-working career- type jobs?  Especially jobs for which many older workers may be qualified 13

14.  Support from SSA  Research assistance and code from Peter Hudomiet and Seth Koch 14