Discussion of
”Debt constraints and Employment"
by Kehoe, Midrigan, and Pastorino
Pedro S. Amaral
Banco de Portugal
Federal Reserve Bank of Cleveland
8th Banco de Portugal Conference on Monetary Economics
11-13 June, 2015
Main Contributions
KMP put some theoretical meat and bones under Mian
and Sufi’s empirical findings
The key drivers are:
Backloaded earnings (through HK growth)
Higher discount rates from tightening debt constraints
Intuition: when earnings come late in one’s career an
increase in discounting lowers expected discounted
surplus disproportionately
Backloading wages
Backloading compensation may be an appealing feature
even outside of HK accumulation:
Burdett and Coles (ECTA 2003): backloading insures
against on the job-search
Deferring compensation to induce effort
If firms have higher discount rates
A perpetual youth economy
Data-like earnings-profile is crucial for argument
Life-cycle issues can be complicated to model
Perpetual youth assumption makes things tractable
No need to keep track of age
But age distributions are off – Mathusalem economy
Labor force experience distribution
0.25
US data
0.2
0.15
0.1
0.05
0
20−24 25−29 30−3435−39 40−44 45−49 50−5455−59 60−64 65−6970−75 75+
Labor force experience distribution
0.25
US data
0.2
Model
0.15
0.1
0.05
0
20−24 25−29 30−3435−39 40−44 45−49 50−5455−59 60−64 65−6970−75 75+
A perpetual youth economy
Does it matter, though?
Model has more experienced people than data
Buchinsky et al. (2010) obtain returns to experience that
are the highest in the literature
As a consequence, the earnings-age profile is still steep
well into one’s grave years
Earnings-age profile
9
8
7
6
5
4
3
2
1
0
50
100
Years of Experience
150
A perpetual youth economy:
consequences
Does it matter, though?
Model has more experienced people than data
Buchinsky et al. (2010) obtain returns to experience that
are the highest in the literature
As a consequence, the earnings-age profile is still steep
well into one’s grave years
This raises two potential problems: calibration and
magnitude of discounting effects
Discounting earnings
Consider discounting the earnings-age profile at different
factors:
a constant one: 0.985, and
a variable one built to replicate the one resulting from the
shock
Discount factor following shock
0.995
0.99
0.985
0.98
0.975
0.97
0.965
0.96
0
2
4
6
Years from shock
8
10
Discounting earnings
Consider discounting that earnings profile at different
factors:
a constant one: 0.985, and
a variable one resulting from the shock
Under perfect foresight, discount profiles differ for
roughly 10 years
Discount factor profiles
1
0.8
0.6
0.4
0.2
0
0
10
20
30
Years from shock
40
50
Discounting earnings: a surprise
The percentage loss in discounted earnings increases
with how old workers were when shock hit
Totally counterintuitive!
Losses in discounted earnings
−0.022
−0.024
−0.026
−0.028
−0.03
−0.032
−0.034
−0.036
−0.038
−0.04
0
50
100
Age when shock hit
150
Discounting earnings: questions
The percentage loss in discounted earnings increases
with how old workers were when shock hit
Totally counterintuitive!
This may be problematic as the old are over-represented
The mean loss in the model is -3.2% and the mean loss
under a replica of the U.S. labor force distribution is -3%.
Discounting earnings: questions
But that still leaves us with two questions:
1
How come the intuition that losses from an increase in
the discount rate are higher under increasing wage
profiles is incorrect?
It is correct for constant discount rates
Discount factor profiles
1
β=0.985
Discount factor
0.8
β=0.96
0.6
0.4
0.2
0
0
20
40
60
80
100
0
Constant earnings
−0.2
% loss
Linear earnings
−0.4
−0.6
−0.8
0
20
40
60
Quarters after shock
80
100
Discounting earnings: questions
But that still leaves us with two questions:
1
How come the intuition that losses from an increase in
the discount rate are higher under increasing wage
profiles is incorrect?
It is correct for constant discount rates
But not for time varying ones
Discount factor profiles
Discount factor
1
0.8
0.6
0.4
0.2
0
20
40
60
80
100
80
100
0.05
Constant earnings
0
% loss
Linear earnings
−0.05
−0.1
0
20
40
60
Quarters after shock
Discounting earnings: questions
But that still leaves us with two questions:
1
How come the intuition that losses from an increase in
the discount rate are higher under increasing wage
profiles is incorrect?
It is correct for constant discount rates
But not for time varying ones
2
Where are the results coming from then? I conjecture
that from changes to the wage policy (and the
consequent changes to surpluses)
Calibration
Why calibrate to employment to working-age
population? You have a counterfactually low steady-state
job-finding rate
Given model demography and your average wage growth
target, your early wage growth is too high
Buchinsky et al (2010) have estimates by education level.
The coefficients for experience vary substantially. Not
clear what is used.
Floden and Linde’s (2001) estimates for the standard
deviation of temporary shocks are yearly
Concern that estimates of standard deviation of initial
log wages not purged from sources of variation not
present in the model.
Consumption (HP)
0.02
0.015
0.01
0.005
0
−0.005
−0.01
−0.015
−0.02
−0.025
2004
2006
2008
2010
2012
2014
2016
Employment to WAP (HP)
0.02
0.015
0.01
0.005
0
−0.005
−0.01
−0.015
−0.02
−0.025
2004
2006
2008
2010
2012
2014
2016
Employment rate (HP)
0.02
0.015
0.01
0.005
0
−0.005
−0.01
−0.015
−0.02
−0.025
2004
2006
2008
2010
2012
2014
2016
Wage policy
3.5
3
Wages
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
z
3
3.5
4
Distributions
Employed
200
150
100
50
0
−1
0
1
2
3
4
5
Unemployed
200
150
100
50
0
−3
−2
−1
0
1
z
2
3
4