Quantitative Modelling for
Decumulation Phase
Zili Zhu, Thomas Sneddon, Colin O’Hare, Bernard Casey
Pavel Shevchenko, Xiaolin Luo, Chenming Bao, Peter Toscas
The Research Project and Its Objectives
Multi-Disciplinary Team
10 people, CSIRO Sydney, Melbourne sites, Monash University,
Warwick University.
Diverse Experience
Behavioral economist, social scientist, actuaries (FIA FIAA),
statisticians, data mining specialists, financial engineers,
Operations Researchers.
Main Deliverables
•
Prototype software for partners/stakeholders.
Retirement products in Australia: limited solutions
The needs of retired individuals:
•
long-term financial security: uncertainty about expenditure, longevity.
•
adequate funds for emergency expenditure: health and retirement home
cost.
Main Focus of this project:
1. To establish what individuals are currently doing in retirement with their funds: SMSF and super balances, spending pattern (from ATO/DSS/DHS data).
2. To determine the probability of ruin before death given fund balances at retirement and annual withdrawal patterns. Annual death probability is also modelled.
3. To design an optimal investment strategy for meeting retirement objectives (e.g. to maximise the fund balance at a particular age with x% probability of ruin or to
minimise the probability of fund ruin before death).
4. To understand retirees behaviour and preferences for new and innovative products, such as annuity products.
Other Focuses of this project
1. Modelling of potential solutions to increase consumer demand for annuities (e.g. compulsory purchase of immediate or deferred lifetime annuities at a particular age) and investment strategies incorporating annuity purchases (e.g. staged purchase of annuities within a portfolio).
2. To determine the % of an individual’s salary contribution to super funds in order to limit their probability of ruin.
3. The probability of ruin of an entire super fund consisting of multiple members,
regarding expected withdrawal/contribution patterns, and expected entry/exit patterns of the fund’s members.
4. Management of liquidity risk in super funds if the component of infrastructure investment is large.
5. Optimal fund investment strategy to achieve objectives regarding probability of ruin– this question may also address the optimal fund strategy given past returns within the fund and their effect upon membership levels of the fund.
Project Progress at a Glance
1. Establish a multi-factor cascade model for future investment
returns and retirement expenditures
•
SUPA (Simulation of Uncertainty for Pension Analysis) model
is completed.
2. Pricing Model for Annuity with Optimal Withdraws and GMDB
•
Prototype pricing software for annuity with GMWB & GMDB
is completed.
3. Incorporate fund holder behavior and forecast ruin years
•
Lee-Carter model for mortality rate
is completed.
•
Impact of individual health risk factors on mortality can be analysed.
completed
•
ATO and DSS (Centrelink) Data re withdrawal behavior
in progress.
•
APRA Data re fund entry/exit and liquidity risk
in progress.
4. Optimal-Decisions-For-Better-Retirement
in progress
•
Use Real-option methodology to construct dynamic optimal portfolios for
retirees
in progress.
Objectives of this workshop
1. To have in-depth feedback
•
Can the project output be used directly for your organisation?
•
What new activities should be pursued in this project?
2. To plan for
•
SUPA model and prototype pricing software for annuity products are made
available to stakeholders.
•
Prototype software for Optimal-Decisions-For-Better-Retirement is made
available.
RECAP: Using SUPA model for simulation
The SUPA model projects long-term
economic behaviour: applies
established stochastic asset model (Wilkie model)
Output from SUPA is used for
Monte Carlo simulation: any number of
paths of desired time span
Additional models extend SUPA:
models for withdrawals, death
RECAP: answer questions through SUPA in
decumulation phase
Member’s questions:
• Probability of ruin of individual’s super (before x years or before death) • $ needed in super fund at retirement (to last for x years or until death) • % of salary that should be contributed to fund to achieve some objective • Optimal investment strategies to reach $ amount or avoid ruin
• Annuity-related purchase/investment strategies
Fund level questions:
• Probability of ruin of an entire group fund • $ needed in fund at given time (liquidity risk)
• % aggregate fund balance that fund should collect • Optimal investment strategy (from fund mgmt POV)
RECAP: Predicting retirement fund for super
contribution rates: 9%, 12%, 15%
Average fund length
(yrs) 15.0992 22.1634 28.7137 Std deviation of fund
length (yrs) 7.9704 10.8315 11.7161 Starting Age 25 25 25 Starting Salary (grows
with wage inflation) $54,425 $54,425 $54,425 Contribution rate 9.00% 12.00% 15.00%
Assumptions
:•person beginning career at 25 •starting salary of $54,425 growing
with wage inflation.
•post-retirement withdrawal
requirement of $42,254 growing with price inflation.
•fund allocation set in accordance
with APRA average asset allocation 2009 data (bottom-right table)
Input variables:
•commencement age •starting salary
•withdrawal requirement •fund allocation
Starting annual expense $42,254.00
Asset Allocation Australian Equities 35.00% International Equities 25.00% Domestic Bonds 12.00% International Bonds 8.00% Cash 20.00%
Average fund length 24.0725
Std deviation of fund length 11.85156064
Starting Age 25
Starting Salary $54,425
RECAP: Probability Distribution of Super Fund Balances
250 paths are shown in
the graph.
SUPA model can project:
oRuin age.
oSuper fund balance by age.
Age 65 70 75 80 85 90 95 100 105
% paths with positive balances 100% 98.67% 69.17% 34.77% 17.52% 9.30% 5.64% 3.63% 2.56%
250 paths for super fund balance across time
$0.00 $500,000.00 $1,000,000.00 $1,500,000.00 $2,000,000.00 $2,500,000.00 $3,000,000.00 $3,500,000.00 $4,000,000.00 $4,500,000.00 $5,000,000.00 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 Series1 Series2 Series3 Series4 Series5 Series6 Series7 Series8 Series9 Series10 Series11 Series12 Series13 Series14 Series15
Example: combining death model with SUPA projection to
show super exhausted before death for a 25-year old
APRA super account balances as “returns”
from ATO Data (by age)
SMSF account balances as “returns” from ATO
Data (by age)
Optimal Decisions For Better Retirement
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 65 70 75 80 85 90 95CPI Domestic Equity Idx Int Equity Idx
0.00% 20.00% 40.00% 60.00% 80.00% 100.00% 65 70 75 80 85 90
Domestic Equity Int Equity Accumulated % of withdraw
Left: 30 year economic scenario generated by SUPA model (each standardised to 0.15 at t=0)
oRed = Domestic equity index value
oGreen = International equity index value
oBlue = Consumer price index value
Dynamic optimal withdrawal/investment decision for portfolio of domestic /international equities given the above scenario.
oRed = Allocation to domestic equities
oGreen = Allocation to international equities
oBlue = Accumulation of withdrawal proportion
Possible Inclusion of: Life cycle utility modelling
Re pr oduc ed fr om J. Di ng (2 01 3)•Impact on retirees’ reactions to changes
in pension policy
•Forecast future costs of the Age Pension
in Australia
•Reference: Paper by Jie Ding of Macquarie University
•Based on utility parameters estimated from empirical data to gauge
retirees’ preference for bequest, housing, luxury consumption.
RECAP: Pricing Variable Annuities
Prototype pricing software has been completed
o to price Variable Annuities with Guaranteed Minimum Withdraw
Benefit (GMWB) and Guaranteed Minimum Death Benefit (GMDB)
Fast numerical evaluation of Variable Annuities
owith different guarantee features and policyholder behaviors
(passive and optimal behaviors)
Pricing model can be distributed to design new annuity
products.
10 60 110 160 210 260 310 4 6 8 10 12 14 16 Fa ir F ees (b p)Minimum Guarantee Withdrawal Rate (%)
Fixed withdrawals
Optimal withdrawals (10% penalty) Optimal withdrawals (5% penalty)
15 20 25 30 35 40 45 50 4 6 8 10 12 14 16 Q ua rt er ly P ay m en t ( bp)
Minimum Guarantee Withdrawal Rate (%)
Combined GMWB and GMDB Buying separate life insurance