The Consumption and Wealth Effects of an
Unanticipated Change in Lifetime Resources
Tullio Jappelli
Università di Napoli Federico II
Mario Padula
Università della Svizzera Italiana
The Bank of Italy’s Analysis of Household Finance, December 3-4, 2015
1
The old days
Do people respond to unanticipated
changes in resources?
1. Nature of change



Fiscal vs. monetary shocks
Temporary (tax rebate, bonus) vs. permanent tax reform
Current vs. lifetime resources
2. Nature of response:
Consumption, wealth, labor supply, portfolio choice
3. Heterogeneity of response
Due to horizon, liquidity constraints, precuationary saving, behavioral
theories
4
Outline
- The reform of severance pay of Italian public employees
- Simulated effect on wealth and consumption.
- Difference-in-difference estimates using SHIW data.
- €1 reduction in severance pay increases wealth-income ratio
by 0.32 and reduces consumption-income ratio by 0.03.
5
Many MPCs…
Consumption response
Coupons
Payroll
Recession
Age
Check in
Asset liquidity
the mail
Debt
Context
Anticipated income
changes
Unanticipated income
changes
Size
Small
Anticipated
increase
Anticipated
decline
Large
Permanent
shock
Positive
Transitory
shock
Negative
6
Measuring MPC from unanticipated income shocks
To estimate response of consumption to income shocks
1.
Write income process (say transitory/permanent decomposition).
Then estimate MPCs using restrictions that theory imposes on the
var-cov matrix of consumption and income growth residuals.

2.
Hall and Mishkin (1982), Blundell, Pistaferri & Preston (2008)
Subjective expectations: how will you spend hypothetical income
increase / decrease? Saving, Consumption, Debt.
- Don’t need data on consumption or worry about income process.
- Can easily look at MPC heterogeneity
Shapiro & Slemrod (various years), Jappelli & Pistaferri (2014)
3.
Natural experiments
Gruber (1997), Browning & Crossley (2001), Paxson (1993), Fuchs-Schundeln (2005),
Di Maggio et al. (2014), Surico &Trezzi (2015).
7
A permanent income change: reform of severance
pay of Italian public employees

Intertemporal model: effect on wealth and consumption.

Difference-in-difference using SHIW data from 1989-2010.

€1 reduction in severance pay reduces consumption
relative to income by 0.03. The wealth-income ratio
increases by 0.32. The offset ratio is 0.4.

Young vs old employees, more than 1 public employee

Robustness
Why is the reform interesting?

Vast literature on the effect of transitory shocks on
consumption, much less on the effect of permanent shocks.

Some papers look at change in disability (Browning and
Crossley, 2001).

Others look at consumption and wealth effects of social
security reforms (Attanasio and Brugiavini, 2003; Attanasio
and Rohwedder 2003; Bottazzi, Jappelli and Padula, 2006).

We study effect on both wealth and consumption looking at
sizable and unanticipated changes in future income.
The severance pay reform

In 2000 Italy replaced its traditional system of severance
pay for public employees with a new system.

Old regime: severance pay was proportional to the final
wage before retirement.

New regime: proportional to lifetime earnings.

The reform entails substantial losses for future
generations of public employees, in the range of
€20,000-30,000, depending on seniority.
Contracts
Severance payment
Private
employees
All
Years of contributions ×0.0691×
yearly salary. Contributions are
capitalized with accrual rate equal
to 0.015+0.75
Public
employees
Pre-reform
All
Years of contributions × 0.80 ×
(final yearly salary / 12)
Public
employees
Post-reform
Before 2000 Pro-rata regime, with two
components, until and after 2010,
with weights given by years of
service.
After 2000
Same as private employees
The size of the shock
Before the
reform
(1)
After the reform
Contracts
Contracts
signed before
signed after
December 2000 December 2000
(2)
(3)
g=1.53%,
y0=15,800
76,195
69,303
58,065
g=2.23%,
y0=18,000
116,517
100,976
77,996
g=2.62%,
y0=20,000
146,234
124,342
92,980
Assumptions: 40 years of work; in column (2) contract is signed in 1995;
g and y are historical averages for all emplyees
Simulations
T 1
max E0   tU (Ct )
Income process
T 1
Yt 1  Pt 1Vt 1
t 0
Ct
N 1
Yt
S
 Rt   Rt  RN
t 0
Pt 1  GPZ
t t 1
t 0
In the pre-reform regime, severance pay is: S  0.8  N  YN 1
N 1
N t
In the post-reform regime: S  0.0691  Yt (1   )
t 0
Results differ depending on when the reform occurs in
the course of the life-cycle.
Wealth-income ratio before and after the
severance pay reform
Change in the wealth-income ratio
c/y before and after the severance pay reform
Change in c/y
An unanticipated negative income shock to lifetime resources
reduces consumption and increases wealth relative to income.
Both effects depend on the size of the shock, and are stronger for
younger workers.
Data

Pooled 1989-2010 sample from SHIW

Age 20-55, 39% public, 61% private

Exclude self-employed and workers near to retirement.
2
3
4
5
6
Wealth-income ratio, by occupation
1990
1995
Public employees
2000
2005
Private employees
2010
.7
.75
.8
.85
.9
Consumption-income ratio, by occupation
1990
1995
Public employees
2000
2005
Private employees
2010
Difference-in-difference estimates
yit Mi  POSTt dMi POSTt xit it
y = wealth or consumption-income ratio
d>0 in the regression for the wealth-income ratio
d<0 in the regression for the consumption-income ratio.
Assumptions
 the reform is exogenous with respect to consumer
decisions;
 the reform is exogenous with respect to changes in
sample composition (no labor supply response).
Baseline estimates
Public employee
Post-reform period
Public employee  post-reform
Age
Male
Family size
College degree
High school diploma
Resident in the Centre
Resident in the South
Wealth-income ratio
Consumption-income
ratio
0.015
(0.065)
0.772
(0.063)***
0.321
(0.101)***
0.088
(0.003)***
0.093
(0.063)
0.051
(0.021)**
1.814
(0.079)***
1.290
(0.052)***
0.507
(0.063)***
-0.040
(0.057)
0.002
(0.010)
0.059
(0.009)***
-0.030
(0.015)**
-0.004
(0.000)***
-0.009
(0.009)
-0.009
(0.003)***
-0.137
(0.012)***
-0.082
(0.008)***
0.037
(0.009)***
0.092
(0.008)***
Regressions by number of public employees
One public employee
More than one public employee
Post-reform period
One public employee  post-reform
More than one public employee
post-reform
Wealth-income
ratio
-0.001
(0.068)
-0.204
(0.101)**
0.745
(0.068)***
0.281
(0.105)***
0.363
(0.161)**
Consumptionincome ratio
-0.016
(0.010)*
-0.090
(0.015)***
0.058
(0.010)***
-0.026
(0.015)*
-0.030
(0.024)
Strongest impact for households with more than 1 public employee
Young vs. old workers
Wealth-income ratio
Public employee
Post-reform
Public employee
 post-reform
Consumption-income ratio
30 years of
contributions
>30 years of
contributions
30 years of
contributions
>30 years of
contributions
-0.078
(0.063)
0.672
(0.071)***
0.406
(0.119)***
0.392
(0.150)***
1.194
(0.152)***
-0.081
(0.253)
0.006
(0.007)
0.067
(0.016)***
-0.034
(0.014)**
-0.022
(0.010)**
0.018
(0.010)*
-0.011
(0.016)
Strongest impact for young public employees.
Robustness tests


Group-specific pre-treatment trends.
Restrict sample to years before the reform and redefine the
post-reform dummy as 1 after 1995.
Add to baseline specification post-1995 dummy and interaction
with public employee dummy.

Define treatment group as households whose all members
are public employees and control group as households whose all
members are private employees.

Regional dummies, sector dummies.

Stronger effects among household with higher education.
Summary

Consumption and wealth effects of 2000 severance
pay reform: unanticipated negative shock to lifetime
resources.

Baseline: on average, reduction in severance pay
equal to one year’s income increases wealth by 4
months income. The offset ratio is 0.4.

The reform reduces consumption by 3pp relative
to income.

Heterogeneity: wealth and consumption response
stronger among households with more than one
public employee and young workers, who expect the
strongest decline in severance pay.
Summary

There is no single MPC.

Identify nature of income shock. Several
strategies: structural models, natural
experiments, direct survey questions.

Effect of shocks depends on whether it is
perceived as permanent or transitory

Importance of heterogeneity of response.