How do Subjective Longevity Expectations Influence Retirement Plans?
M. R. Kahn, M.S. Rutledge, A.Y. Wu
July, 2014
Comments: Deb Dwyer
DEBRA SABATINI DWYER
TECHNOLOGY AND SOCIETY
16th Annual Meeting of the Retirement Research Consortium “Social Security and the Retirement Income System”
National Press Club, Washington, DC
RRC 2014 – Slide 3
Aims to contribute to the literature related to: Retirement Planning and Social Security Policy Rational Expectations and Retirement Decisions
BY: Examining how individuals incorporate Subjective
Life Expectancy (levels and changes) into the planning process
How this translates into Actual Retirement
POTENTIAL CONTRIBUTION
RRC 2014, Slide 4
Large literature on factors influencing retirement Relevance to policies aimed at addressing solvency and
sustainability of SS Programs
Little focus on SLEs Literature does motivate focus on SLEs
Relationship exists but not studied for identification of isolated roleTime Horizon: Important and uncertain resource
Important to understand behavior to set policy to incentivize work (or other goals)
If decision-makers account for longevity in retirement planning they may be more responsive to policies – like extending early and normal age of retirement given rising LE
MERIT OF THE CONTRIBUTION
RRC 2014, Slide 5
“Information is the resolution of uncertainty.” Claude Shannon
Relevance for policy:Analysts have been acknowledging the importance of
accounting for uncertainty in modeling when it comes to retirement
- We learn from actual revealed preferences- Understanding the complexities of how those outcomes
came about, what influences them, is criticalTheories are ambiguous
- This work contributes not only to our understanding of the role of an important resource for retirement (years of life), but the fact that this resource is uncertain.
MERIT OF THE CONTRIBUTION
RRC, 2014, Slide 6
Increasing Life Expectancy has been the basis for discussions of raising eligibility ages for SS OASI benefit take-up
Theory predicts a positive relationship between work life and longevity (as authors acknowledge) because:
Longer life means greater consumption needs Wealth Effect: More time available for work and leisure
Empirical Work is Required to Examine the magnitude of these effects
Depends on preferences (utility weights) over work and leisure as individuals age
BACKGROUND
RRC, Slide 7
Cartoon Upfront: Endogenous Wealth Effect of Life Years:
- Retiring earlier can increase actual/expected life years- Still does not predict a negative relationship between time
horizon and work life necessarily unless…- Strong preferences for a longer time horizon might pose
an investment incentive in the form of retirement- Could explain mixed findings in the literature- Important consideration for interpretation of findings
IS IT UNAMBIGUOUS THEORY?
RRC, 2014, Slide 8
Mixed Findings on role of SLE on Retirement:Hamermesh (1984)Two Surveys – neither directly ask SLERely on national trends, parent mortalityThese indicators do not show a strong relationship with
individual behaviorCannot identify if it is a measurement error issue Even if accurately measured, it is a different cohort
Clearly motivates analysis of SLE on Retirement Behavior as it relates to overall planning for future consumption
-
FIT IN LITERATURE
RRC, 2014, Slide 9
Mixed Findings on role of SLE on Retirement:Evidence on Diminishing Marginal Returns- Hurd et al (2004) findings consistent with DMR theory The value of an additional year is greater to those who have less for
US HRS population – dominating substitution effect of leisure for work Also may retire earlier to gain some years (endogenous wealth effect)
- O’Donnell et al (2008) findings consistent with Wealth Effect Theory
Those with less can afford less retirement (fewer years to spend on both) in England
- Bloom et al (2006) offsetting effects – in US
FIT IN LITERATURE
RRC, 2014, Slide 10
Expectation Formation and Rational Expectations Hamermesh (1985, Expectation, Life Expectancy and
Economic Behavior in QJE) Experimental survey data to capture SLE and how it is formed Skewed toward subjective inputs rather than actual statistics available –
despite the fact that they are aware of the facts Go on to use SLEs to study their role in behavior and utility over work and
leisure
Use Findings to Comment on Need for Social Assistance:“Shortfalls in utility in old age because of skewed or imperfect
forecasting of survival probabilities were found to be relatively small.
This implies that large subsidies to retirees under today's Social Security
system cannot be justified as compensation for an unexpectedly long
retirement for which they failed to save.”
-
FIT IN LITERATURE
RRC, Slide 11
2 studies that examine expected retirementVan Solinge and Henkins (2009) Dutch workerss
Szinovacz et all (2014) HRS – focus on Macroeconomic indicators not SLE
- Neither control for measurement error and endogeneity
FIT IN LITERATURE
RRC, Slide 12
Benitez-Silva et al (2008, Expectations in Micro Data: Rationality Revisited, Empirical Economics)
- Supports Rational Expectation Theory in the formation of SLE and Expected Retirement
- Unlike Hamermesh, after controlling for measurement error bias, model mis-specification, and endogeneity, find individuals on average use information available to predict well
- HRS SLEs are a good indicator of information used in retirement planning
- Use as an instrument for Expected Retirement- Find strong positive relationship between SLE and Expected
Retirement (an additional year of life - delay retirement @ year)- Account for selection in responding to SLE questions
Earlier cohort and focus of work is not on sLE
FIT IN THE LITERATURE CONTINUED
RRC, Slide 13
- More waves – 1992-2010- Thorough and careful analysis- Compare SLE to Objective LE (OLE)- Account for endogenous SLE in expected retirement model- Acknowledge classic measurement error in individual responses and
control for it- Examine CHANGES in SLE to understand how the evolution of
Retirement Planning***- Studies Actual Retirement Behavior to be able to Compare to
Literature- Allow for differences in effects by gender- Sensitivity Checks
SUCCESS IN MAKING CONTRIBUTION
RRC, Slide 14
Sample Selection- Authors do examine this concern for non response in
Expected Retirement- Non-response in indicators of SLE could be a concern
(Benitez-Silva et al, 2008) Sicker, more financially knowledgeable, female, white, married, higher
educated respondents with higher probability of recent parent death are more likely to report
Implications for interpretation of findings – Over-representation of sicker knowledgeable respondents who have given more thought to time horizon
Could exaggerate effect of SLE given omission of folks who have not thought about it as much
Policies effects could be smaller than predicted
COMMENTS/AREAS FOR IMMPROVEMENT
RRC, Slide 15
Most thorough research on SLE and Retirement Planning
- Provides strong evidence that policies aimed at prolonging work life should consider expected longevity – not just actual
- Comment: Consider ambiguous wealth effect in hypotheses formation
- Comment: Consider sample selection regarding SLE responses
Weaker impact on actual behavior – which is ultimately what matters - motivates further research on what drives differences in expected and actual retirement
FINDINGS ARE POLICY RELEVANT