Efficiency versus insurance: The role for fiscal policy in social security privatization
1. Efficiency versus insurance:
The role for fiscal policy in social security privatization
Joanna Tyrowicz (GRAPE, IAAEU, UW and IZA)
Oliwia Komada (GRAPE and WSE)
Krzysztof Makarski (GRAPE and WSE)
SNDE, Dallas, 2019
1 / 39
3. Motivation
• Longevity hazards DB viability: deficit ↑ 1.4% of GDP → reforms
Feldstein, BEA, SSA
• (Any) Reform of social security redistributes
• between generations −→ timing of costs vs gains
• within generations −→ insurance inherent to pension systems features
2 / 39
4. Motivation
• Longevity hazards DB viability: deficit ↑ 1.4% of GDP → reforms
Feldstein, BEA, SSA
• (Any) Reform of social security redistributes
• between generations −→ timing of costs vs gains
• within generations −→ insurance inherent to pension systems features
• Deterministic setup: horse-race between
• efficiency (+)
• fiscal cost of the transition period (–)
2 / 39
5. Motivation
• Longevity hazards DB viability: deficit ↑ 1.4% of GDP → reforms
Feldstein, BEA, SSA
• (Any) Reform of social security redistributes
• between generations −→ timing of costs vs gains
• within generations −→ insurance inherent to pension systems features
• Deterministic setup: horse-race between
• efficiency (+)
• fiscal cost of the transition period (–)
• Stochastic setup: insurance (–)
Nishiyama & Smetters (2007, QJE) and subsequent literature
2 / 39
6. Motivation
• Longevity hazards DB viability: deficit ↑ 1.4% of GDP → reforms
Feldstein, BEA, SSA
• (Any) Reform of social security redistributes
• between generations −→ timing of costs vs gains
• within generations −→ insurance inherent to pension systems features
• Deterministic setup: horse-race between
• efficiency (+)
• fiscal cost of the transition period (–)
• Stochastic setup: insurance (–)
Nishiyama & Smetters (2007, QJE) and subsequent literature
2 / 39
7. Motivation → contribution
Fiscal policy around the pension reform matters:
• counteracts / reinforces redistribution
• affects efficiency (scope of distortions)
3 / 39
8. Motivation → contribution
Fiscal policy around the pension reform matters:
• counteracts / reinforces redistribution
• affects efficiency (scope of distortions)
Literature keeps “low profile” on accompanying fiscal policy
We contribute by
• introducing new fiscal closures: boost efficiency + substitute insurance
• quantifying the role of the insurance motive
3 / 39
11. Literature differs in terms of financing the reform
• Financing within pension system
• contribution rates (20 papers)
Kumru & Thanopoulos (2011, JPE), Bruce & Turnovsky (2013, JPE)
• replacement rate (8 papers)
Boersch-Supan et al. (2014, AER), Kitao (2014, RED)
5 / 39
12. Literature differs in terms of financing the reform
• Financing within pension system
• contribution rates (20 papers)
Kumru & Thanopoulos (2011, JPE), Bruce & Turnovsky (2013, JPE)
• replacement rate (8 papers)
Boersch-Supan et al. (2014, AER), Kitao (2014, RED)
• Financing via fiscal policy
• labor tax (3 papers)
Bouzahzah et al. (2002, JEDC)
• consumption tax (10 papers)
Nishiyama & Smetters (2007, QJE), Diaz-Gimenez & Diaz-Saavedra (2009, RED)
• debt (5 papers )
Song, et al. (2015, AEJ) Lindbeck & Persson (2003, JEL)
5 / 39
13. Literature differs in terms of financing the reform
• Financing within pension system
• contribution rates (20 papers)
Kumru & Thanopoulos (2011, JPE), Bruce & Turnovsky (2013, JPE)
• replacement rate (8 papers)
Boersch-Supan et al. (2014, AER), Kitao (2014, RED)
• Financing via fiscal policy
• labor tax (3 papers)
Bouzahzah et al. (2002, JEDC)
• consumption tax (10 papers)
Nishiyama & Smetters (2007, QJE), Diaz-Gimenez & Diaz-Saavedra (2009, RED)
• debt (5 papers )
Song, et al. (2015, AEJ) Lindbeck & Persson (2003, JEL)
No papers with comparisons across fiscal closures
5 / 39
14. What we do: focus on fiscal policy for a given reform
• Scrutinize links between fiscal policy and pension system reform
• Propose new ways of financing the pensions system reform
• insurance → labor tax progression
• efficiency → capital income gains tax
• timing → smoothing reform cost with public debt
6 / 39
15. What we do: focus on fiscal policy for a given reform
• Scrutinize links between fiscal policy and pension system reform
• Propose new ways of financing the pensions system reform
• insurance → labor tax progression
• efficiency → capital income gains tax
• timing → smoothing reform cost with public debt
Preview of findings
• Nishiyama & Smetters is NOT universal → welfare ↑ in stochastic setup
• Labor tax progression mitigates insurance loss
• Capital income gains tax boost efficiency
6 / 39
16. What we do: focus on fiscal policy for a given reform
• Scrutinize links between fiscal policy and pension system reform
• Propose new ways of financing the pensions system reform
• insurance → labor tax progression
• efficiency → capital income gains tax
• timing → smoothing reform cost with public debt
Preview of findings
• Nishiyama & Smetters is NOT universal → welfare ↑ in stochastic setup
• Labor tax progression mitigates insurance loss
• Capital income gains tax boost efficiency
• Welfare gains and political support only sometimes overlap
• Public debt often “buys” political support
6 / 39
17. OLG model with income shocks, US
Procedure
Baseline: Redistributive PAYG DB
with aging → deficit
7 / 39
18. OLG model with income shocks, US
Procedure
Baseline: Redistributive PAYG DB
with aging → deficit
Reform: Individual DC, 50% funded
• 9 fiscal policies
• 2 pension system adjustments: contribution or benefits
• 7 fiscal closures: tax on labor income, consumption (with or without debt),
+ tax on capital income (w/ or wo/ debt) and progressive income tax
7 / 39
19. OLG model with income shocks, US
Procedure
Baseline: Redistributive PAYG DB
with aging → deficit
Reform: Individual DC, 50% funded
• 9 fiscal policies
• 2 pension system adjustments: contribution or benefits
• 7 fiscal closures: tax on labor income, consumption (with or without debt),
+ tax on capital income (w/ or wo/ debt) and progressive income tax
• government may behave differently in B and R → 81 combinations
7 / 39
20. OLG model with income shocks, US
Procedure
Baseline: Redistributive PAYG DB
with aging → deficit
Reform: Individual DC, 50% funded
• 9 fiscal policies
• 2 pension system adjustments: contribution or benefits
• 7 fiscal closures: tax on labor income, consumption (with or without debt),
+ tax on capital income (w/ or wo/ debt) and progressive income tax
• government may behave differently in B and R → 81 combinations
• compare welfare effect and political support
7 / 39
21. OLG model with income shocks, US
Procedure
Baseline: Redistributive PAYG DB
with aging → deficit
Reform: Individual DC, 50% funded
• 9 fiscal policies
• 2 pension system adjustments: contribution or benefits
• 7 fiscal closures: tax on labor income, consumption (with or without debt),
+ tax on capital income (w/ or wo/ debt) and progressive income tax
• government may behave differently in B and R → 81 combinations
• compare welfare effect and political support
• decompose welfare change into insurance and efficiency
7 / 39
23. Model
Consumers
• uncertain lifetimes: live for 16 periods, with survival rate πj,t < 1
unintended bequest redistributed within a cohort
• uninsurable earnings risk: endogenous labor supply + income shocks
process that follows AR(1) approximated by Markov chain
8 / 39
24. Model
Consumers
• uncertain lifetimes: live for 16 periods, with survival rate πj,t < 1
unintended bequest redistributed within a cohort
• uninsurable earnings risk: endogenous labor supply + income shocks
process that follows AR(1) approximated by Markov chain
• pay taxes (labor, consumption, capital gains) & contribute to pensions
8 / 39
25. Model
Consumers
• uncertain lifetimes: live for 16 periods, with survival rate πj,t < 1
unintended bequest redistributed within a cohort
• uninsurable earnings risk: endogenous labor supply + income shocks
process that follows AR(1) approximated by Markov chain
• pay taxes (labor, consumption, capital gains) & contribute to pensions
• no annuity financial markets with (risk free) interest rate
8 / 39
26. Model
Consumers
• uncertain lifetimes: live for 16 periods, with survival rate πj,t < 1
unintended bequest redistributed within a cohort
• uninsurable earnings risk: endogenous labor supply + income shocks
process that follows AR(1) approximated by Markov chain
• pay taxes (labor, consumption, capital gains) & contribute to pensions
• no annuity financial markets with (risk free) interest rate
Competitive producers
• Cobb-Douglas production function
• capital depreciation rate d
8 / 39
34. Fiscal policy accompanying the reform
• Three new closures details
• progressive labor tax → labor supply
• capital income gain tax (+ debt) → capital accumulation
11 / 39
35. Fiscal policy accompanying the reform
• Three new closures details
• progressive labor tax → labor supply
• capital income gain tax (+ debt) → capital accumulation
• Two closures within pension system details
• contributions → labor supply
• pensions → consumption (of retirees)
11 / 39
36. Fiscal policy accompanying the reform
• Three new closures details
• progressive labor tax → labor supply
• capital income gain tax (+ debt) → capital accumulation
• Two closures within pension system details
• contributions → labor supply
• pensions → consumption (of retirees)
• Four closures outside pension system details
• consumption tax (+ debt) → consumption
• labor tax (+ debt) → labor supply
11 / 39
38. Calibration to replicate US economy (2015)
Preferences
• Preference for leisure φ matches average hours 33%
• Discounting rate δ matches interest rate 4%
12 / 39
39. Calibration to replicate US economy (2015)
Preferences
• Preference for leisure φ matches average hours 33%
• Discounting rate δ matches interest rate 4%
Idiosyncratic productivity shock based on Kruger and Ludwig (2013):
• Persistence η = 0.95
• Variance ση = 0.375
12 / 39
40. Calibration to replicate US economy (2015)
Preferences
• Preference for leisure φ matches average hours 33%
• Discounting rate δ matches interest rate 4%
Idiosyncratic productivity shock based on Kruger and Ludwig (2013):
• Persistence η = 0.95
• Variance ση = 0.375
Pension system
• Replacement rate ρ matches benefits as % of GDP 5.2%
• Contribution rate balances pension system in the initial steady state
• Pension eligibility age at 65
12 / 39
41. Calibration to replicate US economy (2015)
Preferences
• Preference for leisure φ matches average hours 33%
• Discounting rate δ matches interest rate 4%
Idiosyncratic productivity shock based on Kruger and Ludwig (2013):
• Persistence η = 0.95
• Variance ση = 0.375
Pension system
• Replacement rate ρ matches benefits as % of GDP 5.2%
• Contribution rate balances pension system in the initial steady state
• Pension eligibility age at 65
Taxes {τc, τl, τk} match revenue as % of GDP {9.2%, 3.8%, 3.6%}
Depreciation rate d matches investment rate of 25%
12 / 39
42. Calibration to replicate US economy (2015)
Demography is based on UN projections.
number of 20-year-olds mortality rates
13 / 39
44. No reform: deficit ↑
Adjustment in pension parameters
contribution rate ↑ from 7.8% to 9%
tax on pensions ↑ from 0.0% to 17.3%
Adjustment in fiscal parameters
pension system deficit ↑
by 1pp of GDP
14 / 39
45. Funding the reform requires 2% of GDP (temporarily)
capital labor
Pension system deficit temporary ↑ from 0% to 2% of GDP
15 / 39
46. Major effects of the reforming
Pensions linked to contributions
1. reduced labor supply distortion (efficiency ↑)
2. income shocks carry over to retirement (insurance ↓ )
Intuition: fiscal policy can reinforce or attenuate these effects
16 / 39
47. Major effects of the reforming
Pensions linked to contributions
1. reduced labor supply distortion (efficiency ↑)
2. income shocks carry over to retirement (insurance ↓ )
Intuition: fiscal policy can reinforce or attenuate these effects
Eventually taxes decline
(relative to baseline scenario of permanent pension system deficit)
16 / 39
48. Result 1: insurance is small & efficiency is large
capital tax: the highest welfare gain due
to efficiency
progression: the smallest welfare loss
due to insurance
17 / 39
49. Result 2: loss of insurance important but not decisive
τk has larger gain than τc → positive overall welfare effect
18 / 39
50. Result 3: public debt helps gaining political support
Welfare effect – τk
19 / 39
51. Result 3: public debt helps gaining political support
Welfare effect - τk & debt + τk
20 / 39
52. Result 3: public debt helps gaining political support
Welfare effect - transition - τk & debt + τk
actual adjustments
21 / 39
54. Aggregate welfare effects and political support
Fiscal closure Baseline
τk d + τk prog. τ τb τc d + τc τl d + τl
Reform
τk 0.57 0.56 1.01 0.59 0.50 0.65 0.65 0.65 0.66
d + τk 0.54 0.54 0.99 0.56 0.47 0.63 0.63 0.63 0.64
prog. -0.45 -0.45 0.02 -0.13 -0.07 -0.35 -0.35 -0.36 -0.34
τ -0.13 -0.12 0.35 0.09 0.14 -0.03 -0.02 -0.03 -0.01
τb -0.15 -0.14 0.33 0.07 0.13 -0.05 -0.04 -0.05 -0.03
τc -0.14 -0.14 0.33 0.11 0.17 -0.04 -0.03 -0.05 -0.03
d + τc -0.16 -0.16 0.31 0.09 0.15 -0.07 -0.06 -0.07 -0.05
τl -0.46 -0.46 0.01 -0.11 -0.03 -0.36 -0.35 -0.37 -0.35
d + τl -0.45 -0.45 0.01 -0.1 -0.02 -0.36 -0.35 -0.36 -0.35
• τk is always a good idea
• prog. (almost) always better then τl in the reform
22 / 39
55. Aggregate welfare effects and political support
Fiscal closure Baseline
τk d + τk prog. τ τb τc d + τc τl d + τl
Reform
τk 0.57 0.56 1.01 0.59 0.50 0.65 0.65 0.65 0.66
d + τk 0.54 0.54 0.99 0.56 0.47 0.63 0.63 0.63 0.64
prog. -0.45 -0.45 0.02 -0.13 -0.07 -0.35 -0.35 -0.36 -0.34
τ -0.13 -0.12 0.35 0.09 0.14 -0.03 -0.02 -0.03 -0.01
τb -0.15 -0.14 0.33 0.07 0.13 -0.05 -0.04 -0.05 -0.03
τc -0.14 -0.14 0.33 0.11 0.17 -0.04 -0.03 -0.05 -0.03
d + τc -0.16 -0.16 0.31 0.09 0.15 -0.07 -0.06 -0.07 -0.05
τl -0.46 -0.46 0.01 -0.11 -0.03 -0.36 -0.35 -0.37 -0.35
d + τl -0.45 -0.45 0.01 -0.1 -0.02 -0.36 -0.35 -0.36 -0.35
• τk is always a good idea
• prog. (almost) always better then τl in the reform
• little effect of debt on welfare
22 / 39
56. Aggregate welfare effects and political support
Fiscal closure Baseline
τk d + τk prog. τ τb τc d + τc τl d + τl
Reform
τk 0.57 0.56 1.01 0.59 0.50 0.65 0.65 0.65 0.66
d + τk 0.54 0.54 0.99 0.56 0.47 0.63 0.63 0.63 0.64
prog. -0.45 -0.45 0.02 -0.13 -0.07 -0.35 -0.35 -0.36 -0.34
τ -0.13 -0.12 0.35 0.09 0.14 -0.03 -0.02 -0.03 -0.01
τb -0.15 -0.14 0.33 0.07 0.13 -0.05 -0.04 -0.05 -0.03
τc -0.14 -0.14 0.33 0.11 0.17 -0.04 -0.03 -0.05 -0.03
d + τc -0.16 -0.16 0.31 0.09 0.15 -0.07 -0.06 -0.07 -0.05
τl -0.46 -0.46 0.01 -0.11 -0.03 -0.36 -0.35 -0.37 -0.35
d + τl -0.45 -0.45 0.01 -0.1 -0.02 -0.36 -0.35 -0.36 -0.35
• τk is always a good idea
• prog. (almost) always better then τl in the reform
• little effect of debt on welfare
• debt ’buys’ support (for good and bad reforms)
22 / 39
57. Summary
The same reform produces different effects depending on fiscal policy:
• efficiency effects of fiscal policy can be big
• insurance loss can effectively be substituted and is not very large
• smoothing by public debt buys political support
23 / 39
58. Summary
The same reform produces different effects depending on fiscal policy:
• efficiency effects of fiscal policy can be big
• insurance loss can effectively be substituted and is not very large
• smoothing by public debt buys political support
Features of this literature
• Labor has a roughly 10% response to reduced l.m. distortions
do people really “understand” the link between contributions and pensions
• Savings have a roughly 10% reaction to longevity
do people really “understand” the risk of old-age poverty
23 / 39
59. Where we get from here?
Work in progress:
• AIME + OADSI rather than fully redistributive DB
• Continuous income tax progression (e.g. Benabou, 2002)
• Analyze separately effects of longevity and reform
24 / 39
60. Where we get from here?
Work in progress:
• AIME + OADSI rather than fully redistributive DB
• Continuous income tax progression (e.g. Benabou, 2002)
• Analyze separately effects of longevity and reform
Why is capital tax so nice? −→ relatively less responsive to the tax hikes
• Longevity → capital accumulation ↑
• Funded pillar → capital accumulation ↑
24 / 39
61. Where we get from here?
Work in progress:
• AIME + OADSI rather than fully redistributive DB
• Continuous income tax progression (e.g. Benabou, 2002)
• Analyze separately effects of longevity and reform
Why is capital tax so nice? −→ relatively less responsive to the tax hikes
• Longevity → capital accumulation ↑
• Funded pillar → capital accumulation ↑
Preliminary results:
• Longevity is the main driver of changes in the economy
• Capital tax better than consumption tax under longevity
• Without longevity funding and DC reduce welfare regardless of fiscal closure
24 / 39
62. Thank you and
I am happy to take questions!
w: grape.org.pl
t: grape org
f: grape.org
e: j.tyrowicz@grape.org.pl
25 / 39
63. New fiscal closures: capital income gain tax
GO BACK
• capital income gain tax, τk,t
Tt = τl,t(1 − τt)wtLt + τk,trtAt + τc,tCt + Υt
J
j=1
Nj,t
Gt + subsidyt + rtDt = Tt + ∆Dt
• smoothing tax adjustments with public debt
• part of the costs of the reform shifted to the future generations
• fiscal rule
τk,t = (1 − )τfinal
k + τk,t−1 + D
Dt
Yt
−
D
Y
final
• debt in the final steady state the same as in the initial steady state
64. New fiscal closures: labor tax progression
GO BACK
• tr1 the lowest income threshold
• trn is the highest income threshold
• n is the number of income brackets
• m is a tax multiplier such that τi
l,t = τ0
l,t ∗ mi
65. New fiscal closures: labor tax progression
GO BACK
• tr1 the lowest income threshold
• trn is the highest income threshold
• n is the number of income brackets
• m is a tax multiplier such that τi
l,t = τ0
l,t ∗ mi
• Income threshold is multiple of average labor income, (1 − τt)wt
¯lt.
• In the initial steady state m = 1
• In the transition path m = 1.15 and n = 4
66. New fiscal closures: labor tax progression
Total gross labor income (1 − τt)wtLt is a sum of n + 1 components: earnings taxed
by one of n + 1 tax rate.
L
0
t =
¯J
j=1
Nj,t
Ω
min(ωj,t(sj,t)lj,t(sj,t), tr1)dPj,t
L
i
t =
¯J
j=1
Nj,t
Ω
max(min(ωj,t(sj,t)lj,t(sj,t − tr1), tri − tri−1), 0)dPj,t∀i = 1, ..., n
τ
0
l,t =
Gt + subsidyt + ∆Dt − Υ1
J
j=1 Nj,t − τc,1Ct − τk,1rtAt − n
i=0 Li
tτi
l
n
i=0 Li
t
τ
i
l,1 = m
i
∗ τ
0
l,1
67. Fiscal closures within pension system, subsidyt = 0
GO BACK
To keep pension system balanced government may adjust:
• contribution rate τ
• benefits bj (as a tax on benefits)
J
j= ¯Jt
Nj,t(1 − τb,t)bj,t = τt ¯wtLt and subsidyt = 0
68. Fiscal closures outside pension system, subsidyt = 0
GO BACK
• consumption tax, τc,t
• labor tax, τl,t
Tt = τl,t(1 − τt)wtLt + τk,trtAt + τc,tCt + Υt
J
j=1
Nj,t
Gt + subsidyt + rtDt = Tt + ∆Dt
• smoothing tax adjustments with public debt via a fiscal rule: ∀tax ∈ {l, c}
τtax,t = (1 − )τfinal
tax + τtax,t−1 + D (D/Y )t − (D/Y )final
• public debt in the final steady state = the initial steady state
69. Profile of average consumption for τk closure
other closures GO BACK
in line with Gourinchas & Parker (2002, Econometrica)
77. Model solving
GO BACK
• Gauss-Seidel iterative algorithm
• Guess an initial value for k = K/(zL) and compute prices
• Solve individual problem and aggregate it to find new K and L , thus k
• iterate until convergence
78. Model solving
GO BACK
• Gauss-Seidel iterative algorithm
• Guess an initial value for k = K/(zL) and compute prices
• Solve individual problem and aggregate it to find new K and L , thus k
• iterate until convergence
• Consumer problem (backward policy function iterations)
79. Model solving
GO BACK
• Gauss-Seidel iterative algorithm
• Guess an initial value for k = K/(zL) and compute prices
• Solve individual problem and aggregate it to find new K and L , thus k
• iterate until convergence
• Consumer problem (backward policy function iterations)
• implicit tax to reduce state space, Butler (2002)
• policy function iterations with picewise linear interpolation
• within period problem solved with Newton-Raphson
• given initial distribution at age j = 1, transition matrix for idiosyncratic
productivity and the policy functions compute the distribution in any successive
age j.
• aggregation done with Gaussian quadrature
80. Model solving
GO BACK
• Gauss-Seidel iterative algorithm
• Guess an initial value for k = K/(zL) and compute prices
• Solve individual problem and aggregate it to find new K and L , thus k
• iterate until convergence
• Consumer problem (backward policy function iterations)
• implicit tax to reduce state space, Butler (2002)
• policy function iterations with picewise linear interpolation
• within period problem solved with Newton-Raphson
• given initial distribution at age j = 1, transition matrix for idiosyncratic
productivity and the policy functions compute the distribution in any successive
age j.
• aggregation done with Gaussian quadrature
• Transition path, goes between the initial and final steady state