Title: Optimal life-cycle portfolios for heterogeneous workers
1Optimal life-cycle portfolios for heterogeneous
workers
- Fabio Bagliano Giovanna Nicodano
- University of Turin CeRP-Collegio Carlo
Alberto - Carolina Fugazza
- University of Milano Bicocca CeRP-CCA
- 2012 HSE Financial Economics Conference
2Motivation
- The composition of household portfolios respond
to permanent, industry specific labor income
shocks. - We study the response of optimal portfolios to
heterogeneity in correlation and variance of
permanent income shocks in a standard life cycle
model - Cocco Gomes Menhout enriched with risky bonds
- The consensus view (Bodie Merton Samuelson)
- under normal circumstances, investors should
reduce their stock investments as they approach
retirement age - rationale human capital, which decreases as
retirement nears, provides a hedge against
adverse financial outcomes - Problems
- smaller holdings of stocks than predicted
- non participation by the young
- hardly decreasing observed investment profiles
3Our view
- Optimal portfolio share in stocks increases, or
is constant, in age for reasonable parameter
combinations - correlation btw permanent labor income shocks and
stock returns - risk aversion
- variance of income shocks
- Bodie Teussard already find inversion, but for
perfect corrrelation - Rationale for this inversion resolution of
uncertainty regarding social security pension
increases the equity risk bearing capacity as
retirement nears - Non investment in stocks by the young obtains
without participation cost. - At 20 residual uncertainty concerning labor
income is such that the young prefer the bond
market to the stock market, that is more
correlated with labour income
4Implication
- Consensus view inspires Target Date Retirement
Funds default investment rules in DC plans.
These are one-size-fits-all - Vanguard 2045 and 2015 stock allocations of 90
and 57 - Swedish PP 100 in equities until 55, then
gradually into fixed income -
- Our analysis shows that
- Tailored rather than one-size-fits-all portfolio
allocations because of heterogeneity of labor
income shocks and risk aversion - If default is needed, then an equally weighted
portfolio is preferable - TDF scheme delivers very low welfare costs for
standard parameters, but very large ones for
larger background risk
5Previous Literature on Non Decreasing Stock
Profiles
- Benzoni et al (2007) long-run cointegration
between labour income and stock returns - Cocco (2004) presence of housing wealth
- Munk and Sorensen (2010) sensitivity of the
expected labor income growth to the real
short-term interest rate - Here we only have bonds.
- Bonds per se do not alter the consensus view.
- Realistically high correlation and risk aversion
without bonds do not alter consensus view. - Bonds and realistically high correlation and risk
aversion alter consensus view
6Standard life cycle model
- power utility of consumption during life, with
uncertain length - log labour income has a deterministic part, a
temporary shock and a permanent shock, that can
be correlated with stock returns - liquidity constraints prevent from fully insure
against shocks - first pillar social security grants exogenous
replacement ratio after retirement, depending on
last labour income - i.i.d. returns on stocks and risky bonds
correlated with each other - riskless asset
7Calibration (Cocco et al., 2005)
- Base case (black) Variation (red)
- working life 20-65, max age 100, US Mortality
Tables - discount factor b 0.96
- relative risk aversion g 5 and g
8 - Var (permanent shocks) and of se² 0.0106
se² 0.042 - transitory shock to labour income sn²
0.0738 sn² 0.30
- riskless rate rf 0.02
- expected stock and bond risk premia ms 0.04 and
mb 0.02 - standard deviations of asset returns ss0.157
and sb 0.08 - Stock-bond return correlation ?sb 0.2
- Stock-labour income correlation ?sY 0 and
?sY 0.2
8Support for Parametric Assumption
- Observed correlation between permanent labor
income shocks and stock returns - Campbell et al.(2001), Campbell Viceira (2002)
0.33, 0.52 - Heaton Lucas (2000) -0.07, 0.14
- Industry-specific Davis and Willen (2000)
-0.10, 0.40
9Median Investment Profiles Base case
- Insertion of bonds does not alter the age profile
for equities - As in Bodie et al. (1992) and Cocco et al.(2005),
but risky bonds substitute for riskless asset - Prior to retirement, investment in equities is
decreasing in age - The asset allocation of the young is tilted
towards stocks - In the two decades before retirement it gradually
shifts to risky bonds - After retirement, equity share is increasing in
age - As pension wealth is riskless, the retirees
invest in stocks the more so the more financial
wealth is disinvested - Flatter schedule with bequest
10Median Investment Profiles
- As the variance of labour income shocks
increases - no change in the shape of age profiles
- savings and financial wealth increase, lowering
the optimal equity share - this 40 drops to 40 at 40 and keeps relatively
constant until 65
11Median Investment Profiles ?sY 0.2
- The young accumulate stocks more slowly until 25,
since labor income is closer to an implicit
holding of stocks - Then decreasing profile resumes
- At 65 the investor sharply rebalances her
portfolio towards stocks as pension income
becomes certain - high variance both savings and financial wealth
increase, lowering the optimal equity share and
restoring the decreasing profile from age 20 -
12Median Investment ProfilesRRA 8 ?sY 0.2
- workers do not participate when 20-25
- upward sloping age profile for equities
- median equity share never exceeds 0.2 before
retirement -
- higher variance young workers save more and
accumulate larger financial wealth, which leads
to cautious participation in the equity market
13Implications and Evidence on Age Profile for
Equities
- Implication
- Interact risk aversion and correlation to
obtain equity portfolio shares that decrease,
increase or stay constant in age. - Missing interaction may explain divergent results
on empirical relationship - Bodie and Crane (1997) downward sloping
- Heaton and Lucas (2004) horizontal
- Ameriks and Zeldes (2004) increasing or hump
shaped
14Implications and Evidence on Non-Participation
- Implication positive correlation is essential
- Haliassos and Michaelides (2003) not plausible.
- Without bonds, correlation needed to achieve non
participation is 0.5 instead of 0.2 - Early estimates higher correlation for more
educated groups and entrepreneurs, that typically
invest in stocks. - Angerer and Lam (2009) higher correlation for
craftsman, operatives, managers and
administrators, farm laborers, private household
workers and armed forces and education below
college degree.
15Heterogeneity in portfolio shares
- 5th, 50th, 95th percentiles of the
cross-sectional distributions of portfolio shares
conditional on age - decreasing heterogeneity before retirement, when
background risk increases because financial
wealth grows - heterogeneity driven by working histories
(idiosyncratic labour income shocks) together
with low financial wealth to hedge them - more similar optimal investments by workers with
high risk aversion, because of higher financial
wealth and lower heterogeneity
15
16Heterogeneity in portfolio profilesBase Case
normal labor shock variance
high labor shock variance
17HeterogeneityPositive income-stock returns
correlation
high labor shock variance
normal labor shock variance
18HeterogeneityPortfolio shares RRA 8 positive
labor income stock returns correlation (0.2)
19Welfare Costs of Suboptimal Asset Allocation
- Comparison of suboptimal strategies with optimal
one - 1/N strategy of De Miguel et al.
(2008) - Age Rule (100-age) is equally divided between
stocks and bonds - TDF interpolated from observed TDF
- Welfare costs measured in equivalent variation of
lifetime consumption.
20Typical TDF portfolio allocation
21(No Transcript)
22Conclusion
- The optimal portfolio share invested in stock
need not fall in age, even in normal
circumstances - Optimal default investment option ought to be
tied to labour income risk characteristic - Equally weighted strategy better than age rule
and TDF when background risk is high - Current analysis
- Epstein-Zin preferences to investigate driver of
inversion