Did the Chilean Pension Reform Postpone Retirement?

1
Did the Chilean Pension Reform Postpone
Retirement?

• Evidence from Chile
• Alejandra C. Edwards and Estelle James
• Presented at OECD, Paris, 2006

2
Chilean reform of 1981

• Research shows large impact of pension systems on
  retirement age, but no empirical test of DC
• Chilean reform replaced DB with DCcontributions
  go into individual accounts, pension depends on
  accumulation, on actuarial basis
• Old system had incentives to pension early and
  deterrents to work after pensionhigh payroll
  tax, no incremental benefit, work in same job
  restricted
• New system removed these biases and cut payroll
  tax, especially for pensioners who were exempt

3
System as a whole is actuarially fair, but may
not be actuarially fair to each individual

• Some individuals have higher time preference,
  want to save less or start dissaving sooner, want
  to invest in different (riskier?) portfolio, want
  more flexible payouts, expect to die early
• These individuals wont view system as
  actuarially fair to them and still face
  substantial tax component

4
New system should raise lfpr of older workers
thru 2 channels

• Higher retirement age and deterred early pension,
  so liquidity constrained workers workis later
  pensioning due to incentives from actuarial
  fairness or new constraints?
• Increased work propensities among pensionersis
  this due to actuarial fairness or exemption from
  payroll tax, therefore higher net wage?
• We expect smaller change among non-pensionersnot
  exempt from payroll tax and many are not in any
  pension system

5
Pension probabilities among 50-64 fall after 1981
reform (up before)
6
LFP trends reverse after reform(went down
before, up afterwards)
7
particularly among pensioners
8
Aggregate trends are consistent with our
hypotheses.

• Can we isolate the effect of reform from
  variations in individual and macro-economic
  variables? Answer Yes, with some caveats
• We use a difference in differences approach
• We measure marginal behavioral change by cohort
  after reform

9
Data set from Greater Santiago Area Household
Surveys 1960-2004

• Individual level data on
• Labor force participation
• Demographic characteristics
• Labor and pension income
• Shortcomings
• It is not longitudinal, no retrospective data
• But we can build synthetic cohorts--people with
  same birth year, observed at different ages.
• We dont know whether individual is affiliated,
  which system, age pension started, or type
  pension , but we use birth cohort as proxy for
  new system membership

10
Birth cohort proxies new system membership

• Those under age 50 in 1981 were more likely to
  joint new system and fraction in new system is
  larger for each birth year after 1931.
• We define post-reform cohorts as those born after
  1931. Reform effect is predicted to grow with
  each cohort after 1931, until everyone is in the
  new system and has adapted to its incentives.

11
We organize data by cohort

• 31,500 observations men 50-70 years old
• 35 birth cohorts, born 1916-50, pre-reform born
  1916-30 post-reform born 1931-50
• Continuous analysis prob. of new system
  membership proxied by birth year minus 1931
• Discrete analysis 5 or 10 birth years grouped
  and assigned dummies
• COH1 and 2 are pre-reform, born 1916-25
  1926-30, COH3, 4, 5 and 6 are post-reform, born
  1931-35, 1936-40, 1941-45, 1946-50
• we interact dummies with age
• we calculate marginal effect for each cohort
• Interact reform variables with pension status to
  separate impact on pensioners and non-pensioners

12
Probit analysis estimates 1) pension
probabilities and 2) lfp rates as function of

• Individual characteristics measuring the relative
  value of time Age, schooling, marital status,
  presence of young children, nonlabor income
• Time-specific labor market conditions economic
  cycle, unemployment rate, GDP trend
• Lfp rate also a function of pension status,
  amount, pseudo-replacement rate
• Pre and post-reform cohorts and age group
  interactions

13
Results pension prob. by age fall

• Pension rates vary with individual
  characteristics as expected, and increase with
  unemployment
• Pension rates fall if worker reaches 50 after
  reform
• Coh31 is significantly negative for all ages gt 50
• Discrete dummies for each 5-year cohort
• For 50-59, pension rates start falling with COH3
  (46-50 in 1981), accumulated reduction is 12.7
  points by COH6
• This is 67 of starting pension rate for age
  group
• For 60-64, pension rates start falling with COH4
  (41-45 in 1981) by 5.2 points, 15 of starting
  pension rate
• Full cohorts not yet observed total effect
  larger
• For 65-70, pension rates stay the same (few
  observations, no new restrictions apply after age
  65)
• Reason for postponement keeping money in new
  system rewarded, early pension restricted in new
  system

14
Results labor force participation rates of older
workers rise

• Participation rates vary with individual
  characteristics and macro variables as expected
• Pension status, size of pension and
  pseudo-replacement rate have strong negative
  effect on lfpr, as expected. Real pension size
  rose, but not as fast as wages so
  pseudo-replacement rate fell over this period,
  explaining part of aggregate increase
• After all controls, participation rates rise
  sharply for individuals reaching 50 after reform
• (1) more people remain non-pensioners, much
  higher lfpr than pensioners (77 percentage
  points)
• (2) lfpr among pensioners increases
• Participation rates of prime age males and older
  non-pensioners stable

15
Reform effects among pensioners

• In continuous model, interaction of pension and
  coh31 is 1.8 pp per birth year since 1931 for age
  50-65, .6 pp for gt 65
• In discrete cohort model, effect also
  concentrated in pensioners, starts with COH3 or 4
• For ages 50-59, accumulated increase is 20
  pp100 higher than starting lfpr
• For ages 60-65, accumulated increase is 18
  pp125 higher than starting lfpr
• For 65-70, accumulated increase is 5
  ppincomplete
• System is still in transitionpositive marginal
  effect for last cohort observed

16
Macroeconomic effects

• This implies 22 increase in older labor force,
  or 4.4 increase in aggregate labor force and GDP
• Implies growth rate increases .21 per year over
  20 years
• Total increase will be higher once new steady
  state is reached and as in older age groups
  grows

17
Policy implications

• Incentives from shift to new system have had
  positive effects on supply of older workers
• Normal pension age raised, conditions for early
  pension tightened and penalties for postponement
  removed
• Pension growth is actuarially fair with
  contributions
• Pensioners exempted from payroll tax
• Which is more importantactuarial fairness or
  early pension constraints exemption from taxes?
• Large drop in pension prob before 65 not after,
  suggests that early retirement constraints play
  major role
• Large lfp effect among pensioners suggests that
  exemption from payroll tax plays a key role
• Absence of lfpr effect among non-pensioners
  implies that actuarial fairness is less
  important, or that most non-pensioners arent in
  any formal system

18
Future research with new longitudinal data set

• Do workers take pension as soon as eligible?
• Has lfp increased among old system and no-system
  affiliates? What is the differential effect of
  new system, once individual identity is known?
• Do workers near MPG behave differently from other
  workers (different work incentives)