Online Appendix
Unemployment, Participation and Worker Flows
over the Life-Cycle
Sekyu Choi
Universitat Autonoma de Barcelona, Barcelona GSE and MOVE
Alexandre Janiak
Universidad de Chile, Department of Industrial Engineering,
Center for Applied Economics
Benjamın Villena-Roldan
Universidad de Chile, Department of Industrial Engineering,
Center for Applied Economics
1
A Conditional Analysis
We present results of different subsamples of interest. We analysis the relative importance of each
flow for several groups defined by educational attaintment, marital status, and child presence in
the household. The evidence shows a remarkable consistence of baseline results across different
samples.
A.1 Conditional Analysis by Educational Group and Gender
In this subsection, we consider two subsamples: individuals with at most a high school diploma
(ncol) and individuals with at least one year of college education (coll).
A.1.1 Estimated Flows and Stocks by Educational Group and Gender
Figure A1: Life-Cycle Unemployment and Participation Profiles: Males, Non-College vs College
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
U
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P
ncol male coll male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
2
Figure A2: Life-Cycle Unemployment and Participation Profiles: Females, Non-College vs College
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
U
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
P
ncol female coll female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A3: Life-Cycle Profiles of Worker Flows Transitions: Males, Non-College vs College
.01
.02
.03
.04
.05
Pro
b
20 30 40 50 60 70
age
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO.1
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OU
ncol male coll male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
3
Figure A4: Life-Cycle Profiles of Worker Flows Transitions: Females, Non-College vs College
0.0
1.0
2.0
3.0
4
Pro
b
20 30 40 50 60 70
age
EU0
.05
.1.1
5.2
Pro
b
20 30 40 50 60 70
age
EO
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE
.2.2
5.3
.35
.4.4
5
Pro
b
20 30 40 50 60 70
age
UO
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE0
.02
.04
.06
.08
Pro
b
20 30 40 50 60 70
age
OU
ncol female coll female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
4
A.1.2 Markovian Simulations by Educational Group and Gender
Figure A5: Markov-Chain Simulated Unemployment and Participation: Non-College
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.992
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.981
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.988
U females
0.2
.4.6
.8
Pro
b20 30 40 50 60 70
age
R2=0.982
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
5
Figure A6: Markov-Chain Simulated Unemployment and Participation: College
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
R2=0.980
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.988
P males
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
R2=0.949
U females
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
R2=0.986
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
6
A.1.3 Importance Decomposition of Flows by Educational Group and Gender
Figure A7: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Non-College
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.231
EU
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.296
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.020
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.074
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.030
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.810
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
7
Figure A8: AB1C Decomposition of the Importance of Flows: Unemployment, Males, College
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.133
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.688
EO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.066
UE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.061
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.084
OE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.644
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A9: AB1C Decomposition of the Importance of Flows: Participation, Males, Non-College
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.013
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.112
EO.2
.4.6
.81
Pro
b
20 30 40 50 60 70
age
1−R2=0.020
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.177
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.097
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
8
Figure A10: AB1C Decomposition of the Importance of Flows: Participation, Males, College
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.007
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.194
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.015
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.205
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.050
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A11: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Non-College
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.112
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.177
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.038
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.027
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.953
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
9
Figure A12: AB1C Decomposition of the Importance of Flows: Unemployment, Females, College
.04
.06
.08
.1.1
2.1
4
Pro
b
20 30 40 50 60 70
age
1−R2=0.169
EU
.02
.04
.06
.08
.1.1
2
Pro
b
20 30 40 50 60 70
age
1−R2=0.493
EO
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
1−R2=0.031
UE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.089
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.079
OE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.844
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A13: AB1C Decomposition of the Importance of Flows: Participation, Females, Non-College
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.013
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.114
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.020
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.011
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.306
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.125
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
10
Figure A14: AB1C Decomposition of the Importance of Flows: Participation, Females, College
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
EU.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.244
EO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
UO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.332
OE.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.048
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
11
A.2 Conditional Analysis by Marital Status and Gender
In this subsection, we consider married and non-married individuals.
A.2.1 Estimated Flows and Stocks by Marital Status and Gender
Figure A15: Life-Cycle Unemployment and Participation Profiles: Males, Non-Married vs Married.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
U
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P
no married male married male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
12
Figure A16: Life-Cycle Unemployment and Participation Profiles: Females, Non-Married vs Mar-ried
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
U
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
P
no married female married female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A17: Life-Cycle Profiles of Worker Flows Transitions: Males, Non-Married vs Married
0.0
2.0
4.0
6
Pro
b
20 30 40 50 60 70
age
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
OE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
OU
no married male married male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
13
Figure A18: Life-Cycle Profiles of Worker Flows Transitions: Females, Non-Married vs Married
0.0
2.0
4.0
6.0
8
Pro
b
20 30 40 50 60 70
age
EU0
.05
.1.1
5.2
Pro
b
20 30 40 50 60 70
age
EO
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE
.2.2
5.3
.35
.4.4
5
Pro
b
20 30 40 50 60 70
age
UO
.02
.04
.06
.08
.1.1
2
Pro
b
20 30 40 50 60 70
age
OE0
.02
.04
.06
.08
Pro
b
20 30 40 50 60 70
age
OU
no married female married female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
14
A.2.2 Markovian Simulations by Marital Status and Gender
Figure A19: Markov-Chain Simulated Unemployment and Participation: Non-Married
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.972
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.986
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.970
U females
.2.4
.6.8
Pro
b20 30 40 50 60 70
age
R2=0.982
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
15
Figure A20: Markov-Chain Simulated Unemployment and Participation: Married
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
R2=0.963
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.966
P males
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
R2=0.985
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.975
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
16
A.2.3 Importance decomposition of Flows by Marital Status and Gender
Figure A21: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Non-Married
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.262
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.439
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.051
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.139
UO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.094
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.857
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
A.3 Conditional Analysis by Child Presence and Gender
In this subsection, we consider individuals with and without at least a child in their households.
A.3.1 Estimated Flows and Stocks by Child Presence and Gender
17
Figure A22: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Married
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.376
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.227
EO
.05
.1.1
5.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.064
UE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.077
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.083
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.797
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A23: AB1C Decomposition of the Importance of Flows: Participation, Males, Non-Married
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.009
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.104
EO.2
.4.6
.81
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.007
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.262
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.077
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
18
Figure A24: AB1C Decomposition of the Importance of Flows: Participation, Males, Married
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.026
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.085
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.038
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.029
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.171
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.089
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A25: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Non-Married
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.157
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.404
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.036
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.091
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.075
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.878
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
19
Figure A26: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Married
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.043
EU
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.070
EO
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.019
UE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.023
UO
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.031
OE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.498
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A27: AB1C Decomposition of the Importance of Flows: Participation, Females, Non-Married
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.012
EU
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.163
EO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.021
UE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.011
UO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.343
OE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.106
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
20
Figure A28: AB1C Decomposition of the Importance of Flows: Participation, Females, Married
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.014
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.083
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.027
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.208
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.115
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A29: Life-Cycle Unemployment and Participation Profiles: Males, No-Child vs Child
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
U
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P
no child male child male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
21
Figure A30: Life-Cycle Unemployment and Participation Profiles: Females, No-Child vs Child
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
U
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
P
no child female child female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A31: Life-Cycle Profiles of Worker Flows Transitions: Males, No-Child vs Child
.01
.02
.03
.04
.05
Pro
b
20 30 40 50 60 70
age
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OU
no child male child male
Note: Unconditional life-cycle profiles estimated via weighted OLS.
22
Figure A32: Life-Cycle Profiles of Worker Flows Transitions: Females, No-Child vs Child
.01
.02
.03
.04
.05
Pro
b
20 30 40 50 60 70
age
EU0
.05
.1.1
5.2
Pro
b
20 30 40 50 60 70
age
EO
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE
.2.3
.4.5
Pro
b
20 30 40 50 60 70
age
UO
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE0
.02
.04
.06
.08
Pro
b
20 30 40 50 60 70
age
OU
no child female child female
Note: Unconditional life-cycle profiles estimated via weighted OLS.
23
A.3.2 Markovian Simulations by Marital Status and Gender
Figure A33: Markov-Chain Simulated Unemployment and Participation: No-Child
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
R2=0.985
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.971
P males
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
R2=0.984
U females
.2.4
.6.8
Pro
b20 30 40 50 60 70
age
R2=0.967
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
24
Figure A34: Markov-Chain Simulated Unemployment and Participation: Child
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.992
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.984
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.979
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.983
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
25
A.3.3 Importance Decomposition of Flows by Child Status and Gender
Figure A35: AB1C Decomposition of the Importance of Flows: Unemployment, Males, No-Child
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.100
EU
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.159
EO
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.034
UE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.080
UO
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.047
OE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.527
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
26
Figure A36: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Child
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.157
EU
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.288
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.055
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.041
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.478
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A37: AB1C Decomposition of the Importance of Flows: Participation, Males, No-Child
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.020
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.113
EO.2
.4.6
.81
Pro
b
20 30 40 50 60 70
age
1−R2=0.031
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.190
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.098
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
27
Figure A38: AB1C Decomposition of the Importance of Flows: Participation, Males, Child
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.009
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.135
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.194
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.081
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A39: AB1C Decomposition of the Importance of Flows: Unemployment, Females, No-Child
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.047
EU
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.147
EO.0
5.1
.15
.2.2
5.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.034
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.045
UO
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.058
OE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.638
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
28
Figure A40: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Child
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.112
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.195
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.058
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.050
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.842
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A41: AB1C Decomposition of the Importance of Flows: Participation, Females, No-Child
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.023
EU
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.069
EO.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.035
UE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.021
UO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.251
OE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.113
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
29
Figure A42: AB1C Decomposition of the Importance of Flows: Participation, Females, Child
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.013
EU0
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.191
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.019
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.012
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.326
OE0
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.085
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
30
B Effects of the Great Recession
In this Section, we show in greater detail results for the analysis of the Great Recession. To do so,
we report results from our analysis for unconditional transition probabilities, unemployment, and
participation rates before and after January 2007.
B.1 Estimated Flows and Stocks, Before and After 2007
Figure A43: Life-Cycle Unemployment and Participation Profiles: Males, Before and After 2007
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
u male
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
p male
2007−> <−2007
Note: Unconditional life-cycle profiles estimated via weighted OLS.
31
Figure A44: Life-Cycle Unemployment and Participation Profiles: Females, Before and After 2007
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
U female
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
P female
2007−> <−2007
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A45: Life-Cycle Profiles of Worker Flows Transitions: Males, Before and After 2007
.01
.02
.03
.04
.05
Pro
b
20 30 40 50 60 70
age
EU male
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO male.1
.15
.2.2
5.3
Pro
b
20 30 40 50 60 70
age
UE male
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OU male
2007−> <−2007
Note: Unconditional life-cycle profiles estimated via weighted OLS.
32
Figure A46: Life-Cycle Profiles of Worker Flows Transitions: Males, Before and After 2007
0.0
1.0
2.0
3.0
4
Pro
b
20 30 40 50 60 70
age
EU female
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO female
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE female
.2.3
.4.5
Pro
b
20 30 40 50 60 70
age
UO female
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE female
0.0
2.0
4.0
6.0
8
Pro
b
20 30 40 50 60 70
age
OU female
2007−> <−2007
Note: Unconditional life-cycle profiles estimated via weighted OLS.
B.2 Markov Chain Analysis
33
Figure A47: Markov-Chain Simulated Unemployment and Participation: Before 2007
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.994
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.985
P males
0.2
Pro
b
20 30 40 50 60 70
age
R2=0.987
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.985
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A48: Markov-Chain Simulated Unemployment and Participation: After 2007
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
R2=0.980
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.980
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.973
U females
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
R2=0.975
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
34
B.3 Importance decomposition of Flows, before and after 2007
Figure A49: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Before 2007
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.140
EU
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.252
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.021
UE
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.058
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.034
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.657
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
35
Figure A50: AB1C Decomposition of the Importance of Flows: Unemployment, Males, After 2007
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.102
EU
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.308
EO
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.031
UE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.076
UO
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.104
OE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.424
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A51: AB1C Decomposition of the Importance of Flows: Participation, Males, Before 2007
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.009
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.142
EO.2
.4.6
.81
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.195
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.073
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
36
Figure A52: AB1C Decomposition of the Importance of Flows: Participation, Males, After 2007
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.012
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.088
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.022
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.009
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.187
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.099
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
B.4 Role of Inactivity, Before and After 2007
37
Figure A53: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Before2007
0.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.082
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.172
EO
0.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.044
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.045
OE
0.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.874
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A54: AB1C Decomposition of the Importance of Flows: Unemployment, Females, After2007
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.080
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.288
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.028
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.077
UO
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.098
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.616
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
38
Figure A55: AB1C Decomposition of the Importance of Flows: Participation, Females, Before 2007
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.132
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.291
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.083
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A56: AB1C Decomposition of the Importance of Flows: Participation, Females, After 2007
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.018
EU
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.144
EO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.028
UE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.014
UO
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.288
OE
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.099
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
39
Figure A57: AB2 Decomposition Neglecting Inactivity for Life-Cycle Unemployment Rates: Males,Before 2007
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2C, EU,UE mean
1−R2=0.079
.05
.1.1
5.2
.25
Pro
b20 30 40 50 60 70
age
actual AB2C, EU,UE zero
1−R2=0.036
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2F mean, but EU,UE
R2=0.950
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2F zero, but EU,UE
R2=0.985
Note: Unconditional life-cycle profiles estimated via weighted OLS.
40
Figure A58: AB2 Decomposition Neglecting Inactivity for Life-Cycle Unemployment Rates: Males,After 2007
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
actual AB2C, EU,UE mean
1−R2=0.063
.05
.1.1
5.2
.25
.3
Pro
b20 30 40 50 60 70
age
actual AB2C, EU,UE zero
1−R2=0.045
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
actual AB2F mean, but EU,UE
R2=0.921
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
actual AB2F zero, but EU,UE
R2=0.977
Note: Unconditional life-cycle profiles estimated via weighted OLS.
41
Figure A59: AB2 Decomposition Neglecting Inactivity for Life-Cycle Unemployment Rates: Fe-males, Before 2007
0.2
Pro
b
20 30 40 50 60 70
age
actual AB2C, EU,UE mean
1−R2=0.045
.05
.1.1
5.2
.25
Pro
b20 30 40 50 60 70
age
actual AB2C, EU,UE zero
1−R2=0.043
0.2
Pro
b
20 30 40 50 60 70
age
actual AB2F mean, but EU,UE
R2=0.850
0.2
Pro
b
20 30 40 50 60 70
age
actual AB2F zero, but EU,UE
R2=0.965
Note: Unconditional life-cycle profiles estimated via weighted OLS.
42
Figure A60: AB2 Decomposition Neglecting Inactivity for Life-Cycle Unemployment Rates: Fe-males, After 2007
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2C, EU,UE mean
1−R2=0.042
.05
.1.1
5.2
.25
.3
Pro
b20 30 40 50 60 70
age
actual AB2C, EU,UE zero
1−R2=0.055
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2F mean, but EU,UE
R2=0.849
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual AB2F zero, but EU,UE
R2=0.954
Note: Unconditional life-cycle profiles estimated via weighted OLS.
43
C Results for Ages 30 to 50
In this section, we present results restricted for individuals between 30 and 50 years old. Since
the common wisdom seems to be that inactivity matters mostly for younger and older individuals,
restricting our analysis to this population shows how robust our findings are. The fact that annual
probability transitions are close to limit ones (due to large transition probabilities, see section D,
a priori one expect that our results for ages 30-50 to be hardly different from those covering the
baseline range (ages 16-70). Figure A61 shows the results of simulating the unemployment profile
for ages 30 to 50, which gives us comparable accuracy to our baseline results. The results for
participation are slightly less successful.
Figure A61: Markov-Chain Simulated Unemployment and Participation: Age 30-50
.04.
045.
05.0
55.0
6
Pro
b
30 35 40 45 50
age
R2=0.965
U males age 30−50
.88
.9.9
2.9
4
Pro
b
30 35 40 45 50
age
R2=0.981
P males age 30−50
.04.0
45.0
5.055
.06.0
65
Pro
b
30 35 40 45 50
age
R2=0.982
U females age 30−50
.7.7
2.7
4.7
6
Pro
b
30 35 40 45 50
age
R2=0.895
P females age 30−50
actual Markov 30−50
Note: Unconditional life-cycle profiles estimated via weighted OLS.
C.1 Decomposition Exercises, Ages 30 to 50
Below we perform the AB1C decomposition for the range 30 to 50. Figures 21 and 22 show the
results. In line with our previous analysis for the whole age range, the separation probability (EU)
is the most important factor for males, while the OU probability is still important, especially for
workers over 40. In contrast, the job finding probability does not affect the male profile by much.
For females, the most important flow is OU , seconded by the EU and EO probabilities. Hence,
although the importance of inactivity related flows decays in the prime-age group, they are still
important for shaping unemployment profiles. The belief that inactivity flows only matter for
44
younger and older workers is not supported by this evidence.
Figures A64 and A65 show the impact of each particular flow on the participation rate over
ages 30 to 50. For males, the most important flow is the OE, while for females, both EO and OE
are important to explain the profiles.
In sum, while the importance of inactivity related flows is decreased for prime-aged workers,
flows in and out of inactivity still have a substantial influence on unemployment and participation
over the life cycle for this group.
Figure A62: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Age 30-50
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.862
EU
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.060
EO
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.085
UE
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.049
UO
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.102
OE
.04
.045
.05
.055
.06
Pro
b
30 35 40 45 50
age
1−R2=0.289
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
45
Figure A63: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Age 30-50
.04
.045
.05
.055
.06
.065
Pro
b
30 35 40 45 50
age
1−R2=0.120
EU
.04
.045
.05
.055
.06
.065
Pro
b
30 35 40 45 50
age
1−R2=0.094
EO
.04
.045
.05
.055
.06
.065
Pro
b
30 35 40 45 50
age
1−R2=0.030
UE
.04.
045.
05.0
55.0
6.06
5
Pro
b
30 35 40 45 50
age
1−R2=0.017
UO
.04.
045.
05.0
55.0
6.06
5
Pro
b
30 35 40 45 50
age
1−R2=0.036
OE
.04
.045
.05
.055
.06
.065
Pro
b
30 35 40 45 50
age
1−R2=0.140
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
Figure A64: AB1C Decomposition of the Importance of Flows: Participation, Males, Age 30-50
.88
.9.9
2.9
4
Pro
b
30 35 40 45 50
age
1−R2=0.014
EU
.89
.9.9
1.9
2.9
3.9
4
Pro
b
30 35 40 45 50
age
1−R2=0.033
EO.8
9.9
.91
.92
.93
.94
Pro
b
30 35 40 45 50
age
1−R2=0.024
UE
.89
.9.9
1.9
2.9
3.9
4
Pro
b
30 35 40 45 50
age
1−R2=0.022
UO
.89
.9.9
1.9
2.9
3.9
4
Pro
b
30 35 40 45 50
age
1−R2=0.278
OE
.89
.9.9
1.9
2.9
3.9
4
Pro
b
30 35 40 45 50
age
1−R2=0.078
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
46
Figure A65: AB1C Decomposition of the Importance of Flows: Participation, Females, Age 30-50
.71
.72
.73
.74
.75
.76
Pro
b
30 35 40 45 50
age
1−R2=0.061
EU.7
.72
.74
.76
Pro
b
30 35 40 45 50
age
1−R2=0.861
EO
.7.7
2.7
4.7
6
Pro
b
30 35 40 45 50
age
1−R2=0.125
UE
.7.7
2.7
4.7
6
Pro
b
30 35 40 45 50
age
1−R2=0.088
UO
.7.7
2.7
4.7
6
Pro
b
30 35 40 45 50
age
1−R2=0.447
OE.7
.72
.74
.76
Pro
b
30 35 40 45 50
age
1−R2=0.253
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.
47
D Alternative Decomposition Methods
Our decomposition method is similar to the one used by Pissarides (1986) and Shimer (2012).
More specifically, unemployment and labour force participation approximations in the latter are
the result of iterating the Markov chains an infinite number of times. Labour states obtained from
twelve months of transitions (to simulate one year in the life of a worker) with empirical transition
probabilities are not very different from the Markov chain limit. In most cases, the approximation is
accurate so that we can construct theoretical counterparts to the observed proportion of individuals
in each of the three considered states {e, u, o} at age a using the Markov chain limit. Therefore,
the approximation at any age a can be constructed by solving the following linear system1
(EUa + EOa) Ea = UEaUa +OEaOa
(UEa + UOa) Ua = EUaEa +OUaOa
(OEa +OUa) Oa = EOaEa + UOaUa
The interpretation of these equations is straightforward. The left hand side of these equations
represent the flow of individuals transiting away from states {e, u, o} respectively, at the end of age
a. The right hand side accounts for the number of individuals transiting into those same states.
These two numbers must be the same, assuming a stationary age-specific population structure and
stationary transition probabilities xza. Solving for the states, we get functional forms that relate
them to age specific transition rates only.
Ea = E(UEa, UOa, OEa, OUa)
Ua = U(EUa, EOa, OEa, OUa)
Oa = O(EUa, EOa, UEa, UOa)
accordingly, we can construct these “theoretical” counterparts for participation (pa = 1− Oa) and
unemployment rates (ua = Ua/(Ea + Ua)) using the above equations and our estimates of {XZa}:
ua ≈ ua =Ua
Ua + Ea
=OEaEUa +OUa(EUa +EOa)
OEa(UOa + EUa) + UEa(OEa +OUa) +OUa(EUa + EOa)
pa ≈ pa
= 1− Oa =UEa(OEa +OUa) +OEa(UOa + EUa) +OUa(EUa + EOa)
UEaEOa + EOaUOa + UOaEUa + UEa(OEa +OUa) +OEa(UOa + EUa) +OUa(EUa + EOa)
In Figure A66 we plot the observed versus theoretical (constructed) rates, for both men and women.
As seen from the Figure, the theoretical rates follow closely their observed counterparts and pose
a reasonable approximation to the observed profiles. Notice that in order to calculate stocks
1The limiting labour states e, u, o are just the normalized eigenvector (so that its components add up to 1)associated to an eigenvalue of value 1.
48
Figure A66: Limit Markov-Chain Simulated Unemployment and Participation
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=.995
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=.991
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=.99
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=.994
actual lim Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.
of unemployed, employed and inactive workers, the method above does not rely on initial condi-
tions/distribution of workers across employment states but only age-specific transition probabilities.
The goodness of fit of the theoretical rates is due to high monthly transition probabilities, which
dwarfs the effect of initial conditions.
Given that theoretical participation and unemployment rates depend only on age-specific transi-
tion probabilities, we can assess their relative importance in explaining aggregate life-cycle profiles.
Using the same logic as in the “all but one change” (AB1C) method,2 we compute the limiting
states at each age by using our estimates XZa. However, we keep fixed a particular transition
probability at its mean life-cycle value, one at a time, and we allow the rest of them to change
according to age. We present these decompositions for unemployment and participation in Figures
A67 to A70 below.
When comparing the results from this “limit” method to the ones we see in the Markov chain
analysis, we get roughly identical results. In terms of participation, the most important transition
2This is in contrast to what Shimer (2012) does in the context of a business cycle decomposition. He fixes alltransition probabilities at their mean and changes only one, what we labeled the AB1F method above.
49
Figure A67: Limit AB1C Decomposition of the Importance of Flows: Unemployment, Males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual EU at mean
1−R2=0.196
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
actual EO at mean
1−R2=0.445
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual UE at mean
1−R2=0.017
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
actual UO at mean
1−R2=0.060
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
actual OE at mean
1−R2=0.044
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual OU at mean
1−R2=0.821
Note: Unconditional life-cycle profiles estimated via weighted OLS.
probability is the one from employment to inactivity EO. If this transition probability were to be
constant throughout the life-cycle, the participation profile would be flatter. The EO probability is
very important to determine early and late life employment status. Also, movements from inactivity
into the labour force (both OE and OU probabilities) determine to a great extent unemployment
after the age of 60.
As for the life-cycle profile of the unemployment rate, again the EO probability plays an impor-
tant role, followed by the EU as well as the OU transition probabilities. The job finding probability
(UE) does not affect differences in life-cycle participation and unemployment significantly. It turns
out that the transitions into and out from the labour force are quite important in shaping unem-
ployment and participation rates life-cycle profiles.
These results contrasts to Shimer (2012) and Fujita and Ramey (2009) findings in relation to the
business cycle. These authors show that UE and EU flows are enough to account for the cyclical
fluctuations of the unemployment rate. For life-cycle analysis, our evidence shows that inactivity
transitions are key to understand the unemployment and participation by age.
50
Figure A68: Limit AB1C Decomposition of the Importance of Flows: Unemployment, Females
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual EU at mean
1−R2=0.077
0.2
Pro
b
20 30 40 50 60 70
age
actual EO at mean
1−R2=0.338
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual UE at mean
1−R2=0.012
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual UO at mean
1−R2=0.039
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
actual OE at mean
1−R2=0.032
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
actual OU at mean
1−R2=0.984
Note: Unconditional life-cycle profiles estimated via weighted OLS.
51
Figure A69: Limit AB1C Decomposition of the Importance of Flows: Participation, Males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual EU at mean
1−R2=0.018
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual EO at mean
1−R2=0.238
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual UE at mean
1−R2=0.007
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual UO at mean
1−R2=0.016
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual OE at mean
1−R2=0.072
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
actual OU at mean
1−R2=0.009
Note: Unconditional life-cycle profiles estimated via weighted OLS.
52
Figure A70: Limit AB1C Decomposition of the Importance of Flows: Participation, Females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual EU at mean
1−R2=0.013
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual EO at mean
1−R2=0.249
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual UE at mean
1−R2=0.005
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual UO at mean
1−R2=0.011
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual OE at mean
1−R2=0.143
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
actual OU at mean
1−R2=0.013
Note: Unconditional life-cycle profiles estimated via weighted OLS.
53
E Introducing Controls
In this appendix, we present the same figures as in the body of the paper, but here our estimates
include different sets of controls in matrix D below:
fXZatc
√NX
atc =
A∑
a=1
XZaDatc
√NX
atc + βWatc
√NX
atc + ǫatc
√NX
atc (A1)
When we control for time, cohort and state effects, our conclusions remain as in the main body
of the paper. Adding educational attainment dummies to our estimations do not change our results
either, as hinted by the exercise in the paper where we separate samples by educational group.
It is important to consider the interaction between cohort and time effects. In particular, we find
it crucial for isolating the life-cycle component for the post Great Recession period (2007 onwards).
This suggests potentially different effects of the business cycle on different cohorts of the population,
possible due to schooling quality or vintage human capital differences across generations. Once we
properly take these issues into account, our estimated life-cycle profiles are very similar to the
results using unconditional estimates reported in the body of the paper.
E.1 Controlling for Cohort, Time, and State (CTS)
Figure A71: Life-Cycle Unemployment and Participation Profiles, Control CTS
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
U
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P
male female
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
54
Figure A72: Life-Cycle Profiles of Worker Flows Transitions, Males, Control CTS
0.0
1.0
2.0
3.0
4.0
5
Pro
b
20 30 40 50 60 70
age
EU male
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO male
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE male
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE male
0.0
5.1
Pro
b
20 30 40 50 60 70
age
OU male
CI 95% Prob
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
Figure A73: Life-Cycle Profiles of Worker Flows Transitions, Females, Control CTS
0.0
1.0
2.0
3.0
4
Pro
b
20 30 40 50 60 70
age
EU female
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO female
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE female
.2.3
.4.5
Pro
b
20 30 40 50 60 70
age
UO female
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE female
0.0
2.0
4.0
6.0
8
Pro
b
20 30 40 50 60 70
age
OU female
CI 95% Prob
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
55
Figure A74: Markov-Chain Simulated Unemployment and Participation, Control CTS
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.938
U males
.2.4
.6.8
1
Pro
b20 30 40 50 60 70
age
R2=0.956
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.984
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.984
P females
actual Markov
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
56
Figure A75: AB1C Decomposition of the Importance of Flows: Unemployment, Males, ControlCTS
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.329
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.251
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.045
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.144
UO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.026
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.509
OU
actual AB1C, mean
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
57
Figure A76: AB1C Decomposition of the Importance of Flows: Unemployment, Females, ControlCTS
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.066
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.201
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.049
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.060
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.876
OU
actual AB1C, mean
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
Figure A77: AB1C Decomposition of the Importance of Flows: Participation, Males, Control CTS
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.032
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.163
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.046
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.031
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.220
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.105
OU
actual AB1C, mean
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
58
Figure A78: AB1C Decomposition of the Importance of Flows: Participation, Females, ControlCTS
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.011
EU0
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.127
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.312
OE0
.2.4
.6.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.084
OU
actual AB1C, mean
Note: Life-cycle profiles controlled for 5-year cohort effects, 4th-order polynominal cohort-specific time trends, seasonalmonthly dummies, and state effects, estimated via weighted OLS.
59
F Time-Aggregation Results
F.1 Method
Following Shimer (2013) and Elsby, Hobijn, and Sahin (2013), we perform a simple transformation
for each one of matrices Γi, which contain monthly transition probabilities between labour mar-
ket states for each age i. As shown by Shimer (2013), if all eigenvalues of Γi are real, positive
and distinct (as is the case for each considered age in our data), then matrix Γi, containing the
instantaneous transition rates between each labour market state, can be recovered by a simple
eigenvalue/eigenvector transformation: Γi = PiΛiP−1i
, where Pi is the matrix of eigenvectors of
matrix Γi; Λi is the diagonal matrix with eigenvalues of Γi; finally, Λi is the same as Λi, but its
elements are replaced by natural logarithms of the original eigenvalues in Λi. Finally, we use tran-
sition probabilities instead of transition rates, by computing XZ = 1 − exp(−fXZ) where fXZ is
element (X,Z) of matrix Γi. We report monthly time-aggregation corrected transition probabilities
in Figures A79 to A80.
F.2 Robustness of AB1C method:
Given the eigenvalues Λi and the relationship between instantaneous and discrete time transition
probabilities, we can construct transition probabilities for any arbitrary length of time: instead of
month-to-month transitions, we can compute transitions for weeks, days, hours, etc. Note that the
monthly transition probability between states j and k is given by
XZ = 1− exp(−fXZ∆t)
where ∆t = 1 represents one month. Thus, to obtain a weekly, daily or hourly transition probability,
we just have to make ∆t = 7/30, ∆t = 1/30 or ∆t = 1/(24 ∗ 30) respectively. We report the
AB1C decomposition for monthly adjusted transition probabilities in Figures A86 to A85. The
corrected transition probabilities are used to simulate the life-cycle profiles of Unemployment and
Participation of a cohort. While some corrected flows are very different from the unadjusted ones,
the Markovian process accurately predicts stocks of labour statuses only using flows (see Figure
A81).
As we shorten the frequency of transitions, the resulting transition probability matrices (for
example, one per every hour of a 55 year life-cycle in the case of hourly transitions between 16 and
70 years of age) become almost diagonal, since the probability of transiting out of the current labour
force status during the next hour is close to zero. Thus, performing our AB1C decomposition on
these higher frequency transitions gives us an almost ceteris paribus decomposition exercise: since
the relative perturbation to diagonal elements of the matrix when replacing some off-diagonal and
age-specific transition probability with its life-cycle mean becomes minimal. These results are
60
shown in figures A86 to A89 for the hourly case.
Finally, we replicate the AB2 decompositions to understand the effects of omitting inactivity
in the unemployment and participation life-cycle profiles. Using both monthly and hourly time-
aggregation corrected probabilities, we obtain very similar results to those of Section 5 in the main
text. We show the robustness of the AB1C decomposition and confirm the results in the main body
of the paper.
F.3 Monthly Corrected Flows and Simulations
Figure A79: Life-Cycle Profiles of Worker Flows Transitions: Males, Corrected for Time-Aggregation
.01
.02
.03
.04
.05
.06
Pro
b
20 30 40 50 60 70
age
EU male
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO male
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE male
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OU male
raw ta
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
61
Figure A80: Life-Cycle Profiles of Worker Flows Transitions: Females, Corrected for Time-Aggregation
.01
.02
.03
.04
.05
Pro
b
20 30 40 50 60 70
age
EU female
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO female
.15
.2.2
5.3
.35
Pro
b
20 30 40 50 60 70
age
UE female
.2.3
.4.5
.6
Pro
b
20 30 40 50 60 70
age
UO female
0.0
5.1
Pro
b
20 30 40 50 60 70
age
OE female
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OU female
raw ta
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
Figure A81: Markov-Chain Simulated Unemployment and Participation: Corrected for Time-Aggregation
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.981
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.987
P males
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.972
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.988
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
62
F.4 AB1C Decomposition with TA Corrected Probabilities in Various Frequencies
Figure A82: AB1C Decomposition of the Importance of Flows: Unemployment, Males, TA Monthly
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.125
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.251
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.039
UE.0
5.1
.15
.2.2
5.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.064
UO0
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.055
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.591
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
63
Figure A83: AB1C Decomposition of the Importance of Flows: Unemployment, Females, TAMonthly
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.063
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.161
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.049
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.060
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.072
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.822
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
64
Figure A84: AB1C Decomposition of the Importance of Flows: Participation, Males, TA Monthly
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.006
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.132
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.005
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.168
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.115
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
65
Figure A85: AB1C Decomposition of the Importance of Flows: Participation, Females, TA Monthly
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.006
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.146
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.016
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.004
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.235
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.137
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
66
Figure A86: AB1C Decomposition of the Importance of Flows: Unemployment, Males, TA Hourly
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.156
EU
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.289
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.025
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.055
UO
0.0
5.1
.15
.2.2
5
Pro
b
20 30 40 50 60 70
age
1−R2=0.038
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.599
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
67
Figure A87: AB1C Decomposition of the Importance of Flows: Unemployment, Females, TA Hourly
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.092
EU
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.212
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.028
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.047
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.047
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.829
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
68
Figure A88: AB1C Decomposition of the Importance of Flows: Participation, Males, TA Hourly
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.126
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.022
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.005
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.182
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.119
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
69
Figure A89: AB1C Decomposition of the Importance of Flows: Participation, Females, TA Hourly
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.139
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.024
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.006
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.265
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.144
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS.Corrected for Time-Aggregation following Shimer (2013).
70
G Misclassification (MC) Results
G.1 Method
It is well known in the literature that the estimation of transition probabilities from flow data
is sensitive to misclassification (MC) error in recorded labour market states. Given our main
results (i.e., importance of the participation margin for explaining life-cycle unemployment and
participation) any evidence of serious MC error might put in doubt our results, especially since
Unemployment (U) and Out of the labour force (O) are states more likely to be coded with error.3
In this section we analyze the effects of MC error by using two alternative corrections proposed
by the literature: The first approach, follows closely Feng and Hu (2013) (henceforth FH), who use
a latent variable approach to estimate the probability of misclassification of the current true labour
force state, given observed individual histories. The method in FH requires the use of individual
longitudinal information on labour force states, which is available from the CPS since individuals
are followed a total of 8 months (two sets of four consecutive months, separated by an eight month
hiatus). We apply the FH method using labour force histories of three consecutive months, from
where we extract the age-specific joint probability of transiting through specific paths: For example,
given observed information for individuals aged i during months t− 1, t and t+1, we can compute
the set of joint probabilities
Pr(st−1 = j, st = k, st+1 = l)
with {j, k, l} ∈ {E,U,O}. In words, for each age group, we can calculate the fraction who, for
example, transited from employment in period t− 1, to employment in period t and eventually to
unemployment in period t+ 1; i.e., history EEU has a related probability Pr(Et−1, Et, Ut+1) over
all those individuals who have non-missing information for those three consecutive months. The
FH method uses a combination of these probabilities to compute Pr(s∗t |st) with s ∈ {E,U,O} (the
probability of the true labour force state being s∗t given reported state st) through an eigenvalue-
eigenvector decomposition of a suitable arrangement of the above mentioned probabilities into
matrices. Then, using results from Poterba and Summers (1986), one can compute worker flows and
transition probabilities using the true state probabilities. We depart from the FH methodology in
two ways: we use information from three consecutive months, while they use information on months
t− 9, t and t+1. We make this choice in order to minimize data requirements, since we need to be
able to compute these probabilities for each age group: requiring a match between months t− 9, t
and t+1, might produce too much attrition due to the added difficulty of matching workers across
the eight month hiatus in the CPS survey. Our second departure from the FH methodology is also
related to the issue of sample size: we average the resulting age-specific probabilities arising from
3This has been pointed out by Abowd and Zellner (1985) and Poterba and Summers (1986).
71
the FH procedure across 3 ages, in order to minimize the effect of outliers due to low sample size
(specially for younger and older workers). That is, age i corrected stocks and transition probabilities
reflect the average of ages i− 1 to i+ 1.
Second, we follow Elsby, Hobijn, and Sahin (2013) (henceforth EHS), who perform a mechani-
cal recoding of unemployment-nonparticipation cyclers. By matching individuals along the entire
longitudinal dimension of the CPS,4 EHS recode ”obvious” cases of MC: for example, a worker who
in four consecutive months is observed as OOUO (spends the first two months out of the labour
force, the next month as unemployed and the last month out of the labour force) is then recoded
as OOOO. A similar recoding takes place for histories OUOO, UUOU, UOUU and so on.5
Two main differences exist between this procedure and the one from FH: (i) FH require less
data (three matched months instead of four) and (ii) the FH method provides ”corrected” transition
probabilities between all states, not only for those between the unemployment (U) and the out of
the labour (O) states.
Figures A90 and A91 present life-cycle profiles for each transition probability, as they appear
in the main body of the paper (raw), corrected as in Feng and Hu (2013) (FH adj) and corrected
as in Elsby, Hobijn, and Sahin (2013) (EHS adj). Notice that the EHS correction shows differences
for the UO and OU profiles mostly, while the FH adjusts all probabilities downwards. More
importantly for our results, the MC error seems to have a very mild life-cycle component: besides
the last years for OE and OU , the difference between raw and corrected transition probabilities
doesn’t move systematically with age. This can also be seen from figures A94 to A96, where the
probability of being recorded in labour state X given true state X∗ are shown to be very stable over
the life-cycle with one exception: The probability of not being coded as unemployed when truly
unemployed is higher when the worker has less than 20 and more than 65 years of age (inverted
u-shaped of Pr(U |U∗)) which is intuitive, if we think of these stages as the ones where workers are
more ambivalent between participating or not, for example, due to schooling choice decisions for
the young and retirement decisions for the older workers. On the other hand, Pr(O|U∗) is higher
at the beginning and end of the life-cycle, which is the flip side of the previous pattern.
However, when we compute unemployment and participation profiles using the corrected data,
our results differ marginally from those obtained using the raw transition probabilities. The same
applies for our AB1C decomposition exercises for both unemployment and participation, for both
genders. Our interpretation of these results is similar to the one found usually in the literature: at
the aggregate level, these MC errors tend to cancel each other, producing insignificant effects on
flows and rates.
4The method requires four consecutive months of information per worker, as opposed to only two months neededfor the standard flow estimations in the body of the paper.
5See table 2 of Elsby, Hobijn, and Sahin (2013) for a complete list of cycles being recoded.
72
G.2 Monthly Corrected Flows and Stocks
Figure A90: Life-Cycle Profiles of Worker Flows Transitions: Males, Corrected for Misclassification.
0.0
1.0
2.0
3.0
4.0
5
Pro
b
20 30 40 50 60 70
age
EU male
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO male
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE male
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
UO male
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE male
0.0
5.1
Pro
b
20 30 40 50 60 70
age
OU male
raw FH adj EHS adj
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification.
73
Figure A91: Life-Cycle Profiles of Worker Flows Transitions: Females, Corrected for Misclassifica-tion.
0.0
1.0
2.0
3.0
4
Pro
b
20 30 40 50 60 70
age
EU female
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
EO female
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
UE female
.1.2
.3.4
.5
Pro
b
20 30 40 50 60 70
age
UO female
0.0
5.1
.15
Pro
b
20 30 40 50 60 70
age
OE female
0.0
2.0
4.0
6.0
8
Pro
b
20 30 40 50 60 70
age
OU female
raw FH adj EHS adj
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification.
Figure A92: Life-cycle Unemployment Rate Profiles, Corrected for Misclassification.
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
U male
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
U female
raw FH adj EHS adj
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification.
74
Figure A93: Life-cycle Participation Rate Profiles, Corrected for Misclassification.
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P male
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
P female
raw FH adj EHS adj
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification.
Figure A94: Life-Cycle Estimated Misclassification Errors for Employment
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(E|E*)
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(E|U*)
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(E|O*)
male female
Note: Computed on monthly unconditional estimates following Feng and Hu (2013).
75
Figure A95: Life-Cycle Estimated Misclassification Errors for Unemployment
0.2
.4.6
.8P
rob
20 30 40 50 60 70age
Pr(U|E*)
0.2
.4.6
.8P
rob
20 30 40 50 60 70age
Pr(U|U*)
0.2
.4.6
.8P
rob
20 30 40 50 60 70age
Pr(U|O*)
male female
Note: Computed on monthly unconditional estimates following Feng and Hu (2013).
Figure A96: Life-Cycle Estimated Misclassification Errors for Inactivity
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(O|E*)
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(O|U*)
0.2
.4.6
.81
Pro
b
20 30 40 50 60 70age
Pr(O|O*)
male female
Note: Computed on monthly unconditional estimates following Feng and Hu (2013).
76
G.3 Markovian Simulations with MC Corrected Probabilities
Figure A97: Life-Cycle Unemployment and Participation Profiles, Corrected for Misclassificationfollowing Feng and Hu (2013).
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
R2=0.992
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.990
P males
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
R2=0.980
U females
0.2
.4.6
.8
Pro
b20 30 40 50 60 70
age
R2=0.989
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Feng and Hu(2013).
77
Figure A98: Life-Cycle Unemployment and Participation Profiles, Corrected for Misclassificationfollowing Elsby, Hobijn, and Sahin (2013).
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
R2=0.992
U males
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
R2=0.984
P males
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
R2=0.985
U females
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
R2=0.983
P females
actual Markov
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Elsby, Hobijn,and Sahin (2013).
G.4 AB1C Method with MC Corrected Probabilities
78
Figure A99: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Correctedfor Misclassification following Feng and Hu (2013).
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.130
EU
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.234
EO
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.030
UE
0.1
.2.3
.4
Pro
b
20 30 40 50 60 70
age
1−R2=0.055
UO
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.042
OE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.660
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Feng and Hu(2013).
79
Figure A100: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Cor-rected for Misclassification following Feng and Hu (2013).
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.103
EU
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.211
EO
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.034
UE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.052
UO
0.1
.2.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.059
OE
.05
.1.1
5.2
.25
.3
Pro
b
20 30 40 50 60 70
age
1−R2=0.829
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Feng and Hu(2013).
80
Figure A101: AB1C Decomposition of the Importance of Flows: Participation, Males, Correctedfor Misclassification following Feng and Hu (2013).
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.371
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.010
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.015
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.172
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.157
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Feng and Hu(2013).
81
Figure A102: AB1C Decomposition of the Importance of Flows: Participation, Females, Correctedfor Misclassification following Feng and Hu (2013).
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.008
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.271
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.015
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.007
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.254
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.238
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Feng and Hu(2013).
82
Figure A103: AB1C Decomposition of the Importance of Flows: Unemployment, Males, Correctedfor Misclassification following Elsby, Hobijn, and Sahin (2013).
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.184
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.224
EO
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.025
UE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.047
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.036
OE
.05
.1.1
5.2
.25
Pro
b
20 30 40 50 60 70
age
1−R2=0.622
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Elsby, Hobijn,and Sahin (2013).
83
Figure A104: AB1C Decomposition of the Importance of Flows: Unemployment, Females, Cor-rected for Misclassification following Elsby, Hobijn, and Sahin (2013).
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.122
EU
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.158
EO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.022
UE
.05
.1.1
5.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.040
UO
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.044
OE
0.0
5.1
.15
.2
Pro
b
20 30 40 50 60 70
age
1−R2=0.848
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Elsby, Hobijn,and Sahin (2013).
84
Figure A105: AB1C Decomposition of the Importance of Flows: Participation, Males, Correctedfor Misclassification following Elsby, Hobijn, and Sahin (2013).
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.011
EU
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.157
EO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.017
UE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.009
UO
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.203
OE
.2.4
.6.8
1
Pro
b
20 30 40 50 60 70
age
1−R2=0.066
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Elsby, Hobijn,and Sahin (2013).
85
Figure A106: AB1C Decomposition of the Importance of Flows: Participation, Females, Correctedfor Misclassification following Elsby, Hobijn, and Sahin (2013).
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.012
EU
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.153
EO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.018
UE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.011
UO
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.333
OE
0.2
.4.6
.8
Pro
b
20 30 40 50 60 70
age
1−R2=0.075
OU
actual AB1C, mean
Note: Unconditional life-cycle profiles estimated via weighted OLS., Corrected for Misclassification following Elsby, Hobijn,and Sahin (2013).
86