analysis of the salary trajectories in luxembourg...binary data )binary logit distribution censored...

Post on 15-Aug-2020

6 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

includegraphics[height=0.5cm]lsf.eps

Analysis of the salary trajectories in LuxembourgA finite mixture model approach

Jang SCHILTZ (University of Luxembourg)joint work with Jean-Daniel GUIGOU (University of Luxembourg)

& Bruno LOVAT (University Nancy II)

January 19, 2010

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 1 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 2 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 2 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 2 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 2 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 3 / 125

includegraphics[height=0.5cm]lsf.eps

General description of Nagin’s model

We have a collection of individual trajectories.

We try to divide the population into a number of homogenoussubpopulations and to estimate a mean trajectory for each subpopulation.

This is still an inter-individual model, but unlike other classical modelssuch as standard growth curve models, it allows the existence ofsubpolulations with completely different behaviors.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 4 / 125

includegraphics[height=0.5cm]lsf.eps

General description of Nagin’s model

We have a collection of individual trajectories.

We try to divide the population into a number of homogenoussubpopulations and to estimate a mean trajectory for each subpopulation.

This is still an inter-individual model, but unlike other classical modelssuch as standard growth curve models, it allows the existence ofsubpolulations with completely different behaviors.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 4 / 125

includegraphics[height=0.5cm]lsf.eps

General description of Nagin’s model

We have a collection of individual trajectories.

We try to divide the population into a number of homogenoussubpopulations and to estimate a mean trajectory for each subpopulation.

This is still an inter-individual model, but unlike other classical modelssuch as standard growth curve models, it allows the existence ofsubpolulations with completely different behaviors.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 4 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (1)

Consider a population of size N and a variable of interest Y .

Let Yi = yi1 , yi2 , ..., yiT be T measures of the variable, taken at timest1, ...tT for subject number i .

P(Yi ) denotes the probability of Yi

count data ⇒ Poisson distribution

binary data ⇒ Binary logit distribution

censored data ⇒ Censored normal distribution

Aim of the analysis: Find r groups of trajectories of a given kind (forinstance polynomials of degree 4, P(t) = β0 + β1t + β2t

2 + β3t3 + β4t

4.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 5 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)

finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (2)

πj : probability of a given subject to belong to group number j

⇒ πj is the size of group j .

We try to estimate a set of parameters Ω =βj

0, βj1, β

j2, β

j3, β

j4, πj

which

allow to maximize the probability of the measured data.

P j(Yi ) : probability of Yi if subject i belongs to group j

⇒ P(Yi ) =r∑

j=1

πjPj(Yi ). (1)

Finite mixture model(Daniel S. Nagin (Carnegie Mellon University)

)finite : sums across a finite number of groups

mixture : population composed of a mixture of unobserved groups

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 6 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (3)

Hypothesis: In a given group, conditional independence is assumed for thesequential realizations of the elements of Yi !!!

⇒ P j(Yi ) =T∏

t=1

pj(yit ), (2)

where pj(yit ) denotes the probability of yit given membership in group j .

Likelihood of the estimator:

L =N∏

i=1

P(Yi ) =N∏

i=1

r∑j=1

πj

T∏t=1

pj(yit ). (3)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 7 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (3)

Hypothesis: In a given group, conditional independence is assumed for thesequential realizations of the elements of Yi !!!

⇒ P j(Yi ) =T∏

t=1

pj(yit ), (2)

where pj(yit ) denotes the probability of yit given membership in group j .

Likelihood of the estimator:

L =N∏

i=1

P(Yi ) =N∏

i=1

r∑j=1

πj

T∏t=1

pj(yit ). (3)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 7 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (3)

Hypothesis: In a given group, conditional independence is assumed for thesequential realizations of the elements of Yi !!!

⇒ P j(Yi ) =T∏

t=1

pj(yit ), (2)

where pj(yit ) denotes the probability of yit given membership in group j .

Likelihood of the estimator:

L =N∏

i=1

P(Yi ) =N∏

i=1

r∑j=1

πj

T∏t=1

pj(yit ). (3)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 7 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (3)

Hypothesis: In a given group, conditional independence is assumed for thesequential realizations of the elements of Yi !!!

⇒ P j(Yi ) =T∏

t=1

pj(yit ), (2)

where pj(yit ) denotes the probability of yit given membership in group j .

Likelihood of the estimator:

L =N∏

i=1

P(Yi )

=N∏

i=1

r∑j=1

πj

T∏t=1

pj(yit ). (3)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 7 / 125

includegraphics[height=0.5cm]lsf.eps

The Likelihood Function (3)

Hypothesis: In a given group, conditional independence is assumed for thesequential realizations of the elements of Yi !!!

⇒ P j(Yi ) =T∏

t=1

pj(yit ), (2)

where pj(yit ) denotes the probability of yit given membership in group j .

Likelihood of the estimator:

L =N∏

i=1

P(Yi ) =N∏

i=1

r∑j=1

πj

T∏t=1

pj(yit ). (3)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 7 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (1)

Y ∗: latent variable measured by Y .

y∗it = βj

0 + βj1Ageit + βj

2Age2it + βj

3Age3it + βj

4Age4it + εit , (4)

where εit ∼ N (0, σ), σ being a constant standard deviation.

Hence,

yit = Smin si y∗it < Smin,

yit = y∗it si Smin ≤ y∗

it ≤ Smax ,

yit = Smax si y∗it > Smax ,

where Smin and Smax dennote the minimum and maximum of the censorednormal distribution.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 8 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (1)

Y ∗: latent variable measured by Y .

y∗it = βj

0 + βj1Ageit + βj

2Age2it + βj

3Age3it + βj

4Age4it + εit , (4)

where εit ∼ N (0, σ), σ being a constant standard deviation.

Hence,

yit = Smin si y∗it < Smin,

yit = y∗it si Smin ≤ y∗

it ≤ Smax ,

yit = Smax si y∗it > Smax ,

where Smin and Smax dennote the minimum and maximum of the censorednormal distribution.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 8 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (1)

Y ∗: latent variable measured by Y .

y∗it = βj

0 + βj1Ageit + βj

2Age2it + βj

3Age3it + βj

4Age4it + εit , (4)

where εit ∼ N (0, σ), σ being a constant standard deviation.

Hence,

yit = Smin si y∗it < Smin,

yit = y∗it si Smin ≤ y∗

it ≤ Smax ,

yit = Smax si y∗it > Smax ,

where Smin and Smax dennote the minimum and maximum of the censorednormal distribution.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 8 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (2)

Notations :

βjxit = βj0 + βj

1Ageit + βj2Age2

it+ βj

3Age3it

+ βj4Age4

it.

φ: density of standard centered normal law.

Φ: cumulative distribution function of standard centered normal law.

Hence,

pj(yit = Smin) = Φ

(Smin − βjxit

σ

), (5)

pj(yit ) =1

σφ

(yit − βjxit

σ

)pour Smin ≤ yit ≤ Smax , (6)

pj(yit = Smax) = 1− Φ

(Smax − βjxit

σ

). (7)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 9 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (3)

If all the measures are in the interval [Smin,Smax ], we get

L =1

σ

N∏i=1

r∑j=1

πj

T∏t=1

φ

(yit − βjxit

σ

). (8)

It is too complicated to get closed-forms equations

⇒ quasi-Newton procedure maximum research routine

Software:

SAS-based Proc Traj procedureby Bobby L. Jones (Carnegie Mellon University).

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 10 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (3)

If all the measures are in the interval [Smin,Smax ], we get

L =1

σ

N∏i=1

r∑j=1

πj

T∏t=1

φ

(yit − βjxit

σ

). (8)

It is too complicated to get closed-forms equations

⇒ quasi-Newton procedure maximum research routine

Software:

SAS-based Proc Traj procedureby Bobby L. Jones (Carnegie Mellon University).

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 10 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (3)

If all the measures are in the interval [Smin,Smax ], we get

L =1

σ

N∏i=1

r∑j=1

πj

T∏t=1

φ

(yit − βjxit

σ

). (8)

It is too complicated to get closed-forms equations

⇒ quasi-Newton procedure maximum research routine

Software:

SAS-based Proc Traj procedureby Bobby L. Jones (Carnegie Mellon University).

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 10 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (3)

If all the measures are in the interval [Smin,Smax ], we get

L =1

σ

N∏i=1

r∑j=1

πj

T∏t=1

φ

(yit − βjxit

σ

). (8)

It is too complicated to get closed-forms equations

⇒ quasi-Newton procedure maximum research routine

Software:

SAS-based Proc Traj procedureby Bobby L. Jones (Carnegie Mellon University).

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 10 / 125

includegraphics[height=0.5cm]lsf.eps

The case of a censored normal distribution (3)

If all the measures are in the interval [Smin,Smax ], we get

L =1

σ

N∏i=1

r∑j=1

πj

T∏t=1

φ

(yit − βjxit

σ

). (8)

It is too complicated to get closed-forms equations

⇒ quasi-Newton procedure maximum research routine

Software:

SAS-based Proc Traj procedureby Bobby L. Jones (Carnegie Mellon University).

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 10 / 125

includegraphics[height=0.5cm]lsf.eps

A computational trick

The estimations of πj must be in [0, 1].

It is difficult to force this constraint in model estimation.

Instead, we estimate the real parameters θj such that

πj =eθjr∑

j=1

eθj

, (9)

Finally,

L =1

σ

N∏i=1

r∑j=1

eθjr∑

j=1

eθj

T∏t=1

φ

(yit − βjxit

σ

). (10)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 11 / 125

includegraphics[height=0.5cm]lsf.eps

A computational trick

The estimations of πj must be in [0, 1].

It is difficult to force this constraint in model estimation.

Instead, we estimate the real parameters θj such that

πj =eθjr∑

j=1

eθj

, (9)

Finally,

L =1

σ

N∏i=1

r∑j=1

eθjr∑

j=1

eθj

T∏t=1

φ

(yit − βjxit

σ

). (10)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 11 / 125

includegraphics[height=0.5cm]lsf.eps

A computational trick

The estimations of πj must be in [0, 1].

It is difficult to force this constraint in model estimation.

Instead, we estimate the real parameters θj such that

πj =eθjr∑

j=1

eθj

, (9)

Finally,

L =1

σ

N∏i=1

r∑j=1

eθjr∑

j=1

eθj

T∏t=1

φ

(yit − βjxit

σ

). (10)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 11 / 125

includegraphics[height=0.5cm]lsf.eps

A computational trick

The estimations of πj must be in [0, 1].

It is difficult to force this constraint in model estimation.

Instead, we estimate the real parameters θj such that

πj =eθjr∑

j=1

eθj

, (9)

Finally,

L =1

σ

N∏i=1

r∑j=1

eθjr∑

j=1

eθj

T∏t=1

φ

(yit − βjxit

σ

). (10)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 11 / 125

includegraphics[height=0.5cm]lsf.eps

A computational trick

The estimations of πj must be in [0, 1].

It is difficult to force this constraint in model estimation.

Instead, we estimate the real parameters θj such that

πj =eθjr∑

j=1

eθj

, (9)

Finally,

L =1

σ

N∏i=1

r∑j=1

eθjr∑

j=1

eθj

T∏t=1

φ

(yit − βjxit

σ

). (10)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 11 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (1)

Muthen and Shedden (1999): Generalized growth curve model

Elegant and technically demanding extension of the uncensored normalmodel.

Adds random effects to the parameters βj that define a group’s meantrajectory.

Trajectories of individual group members can vary from the grouptrajectory.

Software:

Mplus package by L.K. Muthen and B.O Muthen.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 12 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (1)

Muthen and Shedden (1999): Generalized growth curve model

Elegant and technically demanding extension of the uncensored normalmodel.

Adds random effects to the parameters βj that define a group’s meantrajectory.

Trajectories of individual group members can vary from the grouptrajectory.

Software:

Mplus package by L.K. Muthen and B.O Muthen.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 12 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (1)

Muthen and Shedden (1999): Generalized growth curve model

Elegant and technically demanding extension of the uncensored normalmodel.

Adds random effects to the parameters βj that define a group’s meantrajectory.

Trajectories of individual group members can vary from the grouptrajectory.

Software:

Mplus package by L.K. Muthen and B.O Muthen.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 12 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (1)

Muthen and Shedden (1999): Generalized growth curve model

Elegant and technically demanding extension of the uncensored normalmodel.

Adds random effects to the parameters βj that define a group’s meantrajectory.

Trajectories of individual group members can vary from the grouptrajectory.

Software:

Mplus package by L.K. Muthen and B.O Muthen.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 12 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (1)

Muthen and Shedden (1999): Generalized growth curve model

Elegant and technically demanding extension of the uncensored normalmodel.

Adds random effects to the parameters βj that define a group’s meantrajectory.

Trajectories of individual group members can vary from the grouptrajectory.

Software:

Mplus package by L.K. Muthen and B.O Muthen.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 12 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (2)

Advantage of GGCM

Fewer groups are required to specify a satisfactory model.

Disadvantages of GGCM

1 Difficult to extend to other types of data.

2 Group cross-over effects.

3 can create the illusion of non-existing groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 13 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (2)

Advantage of GGCM

Fewer groups are required to specify a satisfactory model.

Disadvantages of GGCM

1 Difficult to extend to other types of data.

2 Group cross-over effects.

3 can create the illusion of non-existing groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 13 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (2)

Advantage of GGCM

Fewer groups are required to specify a satisfactory model.

Disadvantages of GGCM

1 Difficult to extend to other types of data.

2 Group cross-over effects.

3 can create the illusion of non-existing groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 13 / 125

includegraphics[height=0.5cm]lsf.eps

Muthen’s model (2)

Advantage of GGCM

Fewer groups are required to specify a satisfactory model.

Disadvantages of GGCM

1 Difficult to extend to other types of data.

2 Group cross-over effects.

3 can create the illusion of non-existing groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 13 / 125

includegraphics[height=0.5cm]lsf.eps

Model Selection

Bayesian Information Criterion:

BIC = log(L)− 0, 5k log(N), (11)

where k denotes the number of parameters in the model.

Rule:

The bigger the BIC, the better the model!

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 14 / 125

includegraphics[height=0.5cm]lsf.eps

Model Selection

Bayesian Information Criterion:

BIC = log(L)− 0, 5k log(N), (11)

where k denotes the number of parameters in the model.

Rule:

The bigger the BIC, the better the model!

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 14 / 125

includegraphics[height=0.5cm]lsf.eps

Model Selection

Bayesian Information Criterion:

BIC = log(L)− 0, 5k log(N), (11)

where k denotes the number of parameters in the model.

Rule:

The bigger the BIC, the better the model!

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 14 / 125

includegraphics[height=0.5cm]lsf.eps

Posterior Group-Membership Probabilities

Posterior probability of individual i ’s membership in group j : P(j/Yi ).

Bayes’s theorem

⇒ P(j/Yi ) =P(Yi/j)πjr∑

j=1

P(Yi/j)πj

. (12)

Bigger groups have on average larger probability estimates.

To be classified into a small group, an individual really needs to bestrongly consistent with it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 15 / 125

includegraphics[height=0.5cm]lsf.eps

Posterior Group-Membership Probabilities

Posterior probability of individual i ’s membership in group j : P(j/Yi ).

Bayes’s theorem

⇒ P(j/Yi ) =P(Yi/j)πjr∑

j=1

P(Yi/j)πj

. (12)

Bigger groups have on average larger probability estimates.

To be classified into a small group, an individual really needs to bestrongly consistent with it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 15 / 125

includegraphics[height=0.5cm]lsf.eps

Posterior Group-Membership Probabilities

Posterior probability of individual i ’s membership in group j : P(j/Yi ).

Bayes’s theorem

⇒ P(j/Yi ) =P(Yi/j)πjr∑

j=1

P(Yi/j)πj

. (12)

Bigger groups have on average larger probability estimates.

To be classified into a small group, an individual really needs to bestrongly consistent with it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 15 / 125

includegraphics[height=0.5cm]lsf.eps

Posterior Group-Membership Probabilities

Posterior probability of individual i ’s membership in group j : P(j/Yi ).

Bayes’s theorem

⇒ P(j/Yi ) =P(Yi/j)πjr∑

j=1

P(Yi/j)πj

. (12)

Bigger groups have on average larger probability estimates.

To be classified into a small group, an individual really needs to bestrongly consistent with it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 15 / 125

includegraphics[height=0.5cm]lsf.eps

Posterior Group-Membership Probabilities

Posterior probability of individual i ’s membership in group j : P(j/Yi ).

Bayes’s theorem

⇒ P(j/Yi ) =P(Yi/j)πjr∑

j=1

P(Yi/j)πj

. (12)

Bigger groups have on average larger probability estimates.

To be classified into a small group, an individual really needs to bestrongly consistent with it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 15 / 125

includegraphics[height=0.5cm]lsf.eps

Use for Model diagnostics (2)

Diagnostic 1: Average Posterior Probability of Assignment

AvePP should be at least 0, 7 for all groups.

Diagonostic 2: Odds of Correct Classification

OCCj =AvePPj/1− AvePPj

πj/1− πj. (13)

OCCj should be greater than 5 for all groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 16 / 125

includegraphics[height=0.5cm]lsf.eps

Use for Model diagnostics (2)

Diagnostic 1: Average Posterior Probability of Assignment

AvePP should be at least 0, 7 for all groups.

Diagonostic 2: Odds of Correct Classification

OCCj =AvePPj/1− AvePPj

πj/1− πj. (13)

OCCj should be greater than 5 for all groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 16 / 125

includegraphics[height=0.5cm]lsf.eps

Use for Model diagnostics (2)

Diagnostic 1: Average Posterior Probability of Assignment

AvePP should be at least 0, 7 for all groups.

Diagonostic 2: Odds of Correct Classification

OCCj =AvePPj/1− AvePPj

πj/1− πj. (13)

OCCj should be greater than 5 for all groups.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 16 / 125

includegraphics[height=0.5cm]lsf.eps

Use for Model diagnostics (2)

Diagonostic 3: Comparing πj to the Proportion of the SampleAssigned to Group j

The ratio of the two should be close to 1.

Diagonostic 4: Confidence Intervals for Group MembershipProbabilities

The confidence intervals for group membership probabilities estimatesshould be narrow, i.e. standard deviation of πj should be small.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 17 / 125

includegraphics[height=0.5cm]lsf.eps

Use for Model diagnostics (2)

Diagonostic 3: Comparing πj to the Proportion of the SampleAssigned to Group j

The ratio of the two should be close to 1.

Diagonostic 4: Confidence Intervals for Group MembershipProbabilities

The confidence intervals for group membership probabilities estimatesshould be narrow, i.e. standard deviation of πj should be small.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 17 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 18 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

The data

Salaries of workers in the private sector in Luxembourg from 1940 to 2006.

About 7 million salary lines corresponding to 718.054 workers.

Some sociological variables:

gender (male, female)

nationality and residentship (luxemburgish residents, foreign residents,foreign non residents)

working status (white collar worker, blue collar worker)

year of birth

age in the first year of professional activity

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 19 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Dataset transformations

Mathematica programming

1 row per year → 1 row per worker

selection of the period we are interested in

taking out the years without years up to a maximum of five years

selection of the people who worked the number of years we areinterested in

Transformations in SPSS

elimination of all the workers who had monthly salaries above 15.000

transforming all the salaries above 7.577 to 7.577

creation of the time variables necessary for the Proc Traj procedure

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 20 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons

(Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons (Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons (Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons (Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons (Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Proc Traj procedure

Selection of the time period for macroeconomic reasons (Crisis in the steelindustry and emergence of the financial market place of Luxembourg)

20 years of work for workers beginning their carrier between 1982 and 1987

Proc Traj Macro:

DATA TEST;INPUT ID O1-O20 T1-T20;CARDS;

dataRUN;

PROC TRAJ DATA=TEST OUTPLOT=OP OUTSTAT=OS OUT=OFOUTEST=OE ITDETAIL;

ID ID; VAR O1-O20; INDEP T1-T20;MODEL CNORM; MAX 8000; NGROUPS 6; ORDER 4 4 4 4 4 4;

RUN;

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 21 / 125

includegraphics[height=0.5cm]lsf.eps

Results for 9 groups (1)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 22 / 125

includegraphics[height=0.5cm]lsf.eps

Results for 9 groups (1)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 22 / 125

includegraphics[height=0.5cm]lsf.eps

Results for 9 groups (2)

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 23 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 24 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

13.4 % of the population

P(x) = 590 + 388t − 14t2 + 0.009t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 25 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

13.4 % of the population

P(x) = 590 + 388t − 14t2 + 0.009t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 25 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

13.4 % of the population

P(x) = 590 + 388t − 14t2 + 0.009t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 25 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 26 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 27 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 28 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 28 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 29 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 29 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 30 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 31 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 32 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 33 / 125

includegraphics[height=0.5cm]lsf.eps

1st group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 34 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

16.9 % of the population

P(x) = 785 + 278t − 28t2 + 1.18t3 − 0.016t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 35 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

16.9 % of the population

P(x) = 785 + 278t − 28t2 + 1.18t3 − 0.016t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 35 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

16.9 % of the population

P(x) = 785 + 278t − 28t2 + 1.18t3 − 0.016t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 35 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 36 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 37 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 38 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 38 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 39 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 39 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 40 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 41 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 42 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 43 / 125

includegraphics[height=0.5cm]lsf.eps

2nd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 44 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

20.8 % of the population

P(x) = 709 + 318t − 21.5t2 + 0.62t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 45 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

20.8 % of the population

P(x) = 709 + 318t − 21.5t2 + 0.62t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 45 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

20.8 % of the population

P(x) = 709 + 318t − 21.5t2 + 0.62t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 45 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 46 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 47 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 48 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 48 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 49 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 49 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 50 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 51 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 52 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 53 / 125

includegraphics[height=0.5cm]lsf.eps

3rd group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 54 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

7.9 % of the population

P(x) = 976 + 474t − 29.6t2 − 0.029t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 55 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

7.9 % of the population

P(x) = 976 + 474t − 29.6t2 − 0.029t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 55 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

7.9 % of the population

P(x) = 976 + 474t − 29.6t2 − 0.029t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 55 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 56 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 57 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 58 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 59 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 60 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 61 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 62 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 63 / 125

includegraphics[height=0.5cm]lsf.eps

4th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 64 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

14.9 % of the population

P(x) = 1452 + 490t − 29.6t2 + 1.38t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 65 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

14.9 % of the population

P(x) = 1452 + 490t − 29.6t2 + 1.38t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 65 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

14.9 % of the population

P(x) = 1452 + 490t − 29.6t2 + 1.38t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 65 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 66 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 67 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 68 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 68 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 69 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 69 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 70 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 71 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 72 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 73 / 125

includegraphics[height=0.5cm]lsf.eps

5th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 74 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

4.8 % of the population

P(x) = 2089− 0.017t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 75 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

4.8 % of the population

P(x) = 2089− 0.017t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 75 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

4.8 % of the population

P(x) = 2089− 0.017t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 75 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 76 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 77 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 78 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 79 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 80 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 81 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 82 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 83 / 125

includegraphics[height=0.5cm]lsf.eps

6th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 84 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

6.6 % of the population

P(x) = 2556 + 484t − 29.9t2 + 0.66t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 85 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

6.6 % of the population

P(x) = 2556 + 484t − 29.9t2 + 0.66t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 85 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

6.6 % of the population

P(x) = 2556 + 484t − 29.9t2 + 0.66t3

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 85 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 86 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 87 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 88 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 89 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 90 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 91 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 92 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 93 / 125

includegraphics[height=0.5cm]lsf.eps

7th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 94 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

8.4 % of the population

P(x) = 1987 + 537t − 52.7t2 + 2.06t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 95 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

8.4 % of the population

P(x) = 1987 + 537t − 52.7t2 + 2.06t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 95 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

8.4 % of the population

P(x) = 1987 + 537t − 52.7t2 + 2.06t3 − 0.028t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 95 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 96 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 97 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 98 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 98 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 99 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Men:

Women:

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 99 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 100 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 101 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 102 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 103 / 125

includegraphics[height=0.5cm]lsf.eps

8th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 104 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

6.4 % of the population

P(x) = 3873 + 206t + 30t2− 2.89t3 + 0.06t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 105 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

6.4 % of the population

P(x) = 3873 + 206t + 30t2− 2.89t3 + 0.06t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 105 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

6.4 % of the population

P(x) = 3873 + 206t + 30t2− 2.89t3 + 0.06t4

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 105 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 106 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 107 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 108 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 109 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 110 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 111 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 112 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 113 / 125

includegraphics[height=0.5cm]lsf.eps

9th group

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 114 / 125

includegraphics[height=0.5cm]lsf.eps

Salary Dynamics

Mean annual salary growth:

Group 1 Group 2 Group 3 Group 4 Group 5

λi = 3.07% λ2 = 0.96% λ3 = 1.45% λ4 = 2.82% λ5 = 0.19%

Group 6 Group 7 Group 8 Group 9λ6 = 2.58% λ7 = 1.28% λ8 = 0.48% λ9 = 1.09%

Flat curves : groups 2,5 and 8.

Normal salary growth: groups 3,7 and 9.

Dynamic trajectories: groups 1,4 and 6.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 115 / 125

includegraphics[height=0.5cm]lsf.eps

Salary Dynamics

Mean annual salary growth:

Group 1 Group 2 Group 3 Group 4 Group 5

λi = 3.07% λ2 = 0.96% λ3 = 1.45% λ4 = 2.82% λ5 = 0.19%

Group 6 Group 7 Group 8 Group 9λ6 = 2.58% λ7 = 1.28% λ8 = 0.48% λ9 = 1.09%

Flat curves : groups 2,5 and 8.

Normal salary growth: groups 3,7 and 9.

Dynamic trajectories: groups 1,4 and 6.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 115 / 125

includegraphics[height=0.5cm]lsf.eps

Salary Dynamics

Mean annual salary growth:

Group 1 Group 2 Group 3 Group 4 Group 5

λi = 3.07% λ2 = 0.96% λ3 = 1.45% λ4 = 2.82% λ5 = 0.19%

Group 6 Group 7 Group 8 Group 9λ6 = 2.58% λ7 = 1.28% λ8 = 0.48% λ9 = 1.09%

Flat curves : groups 2,5 and 8.

Normal salary growth: groups 3,7 and 9.

Dynamic trajectories: groups 1,4 and 6.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 115 / 125

includegraphics[height=0.5cm]lsf.eps

Salary Dynamics

Mean annual salary growth:

Group 1 Group 2 Group 3 Group 4 Group 5

λi = 3.07% λ2 = 0.96% λ3 = 1.45% λ4 = 2.82% λ5 = 0.19%

Group 6 Group 7 Group 8 Group 9λ6 = 2.58% λ7 = 1.28% λ8 = 0.48% λ9 = 1.09%

Flat curves : groups 2,5 and 8.

Normal salary growth: groups 3,7 and 9.

Dynamic trajectories: groups 1,4 and 6.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 115 / 125

includegraphics[height=0.5cm]lsf.eps

Outline

1 Nagin’s Finite Mixture Model

2 The Luxemburgish salary trajectories

3 Description of the groups

4 Economic Modeling

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 116 / 125

includegraphics[height=0.5cm]lsf.eps

Dummy example

2 trajectories S1 and S2 with group size 60% and 40% of the population.

Length of the professional life: T = 40 years.

Additional life expectancy: T ∗ = 20 years.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 117 / 125

includegraphics[height=0.5cm]lsf.eps

Dummy example

2 trajectories S1 and S2 with group size 60% and 40% of the population.

Length of the professional life: T = 40 years.

Additional life expectancy: T ∗ = 20 years.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 117 / 125

includegraphics[height=0.5cm]lsf.eps

Dummy example

2 trajectories S1 and S2 with group size 60% and 40% of the population.

Length of the professional life: T = 40 years.

Additional life expectancy: T ∗ = 20 years.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 117 / 125

includegraphics[height=0.5cm]lsf.eps

Dummy example

2 trajectories S1 and S2 with group size 60% and 40% of the population.

Length of the professional life: T = 40 years.

Additional life expectancy: T ∗ = 20 years.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 117 / 125

includegraphics[height=0.5cm]lsf.eps

Hypotheses

Salaries grow linearly, S1 with a starting value of 1500 and a growthcoefficient of 3 %, S2 with a starting value of 1000 and a growthcoefficient of 2 %.

Pensions grow also linearly, S1 with a starting value of 2718 and a growthcoefficient of 2%, S2 with a starting value of 1104 and a growthcoefficient of 1%.

Luxembourg adopts a repartition model, which means that the currentpensions are paid with the tax incomes from the current workers. Eachgeneration hence pays the pension for the generation before it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 118 / 125

includegraphics[height=0.5cm]lsf.eps

Hypotheses

Salaries grow linearly, S1 with a starting value of 1500 and a growthcoefficient of 3 %, S2 with a starting value of 1000 and a growthcoefficient of 2 %.

Pensions grow also linearly, S1 with a starting value of 2718 and a growthcoefficient of 2%, S2 with a starting value of 1104 and a growthcoefficient of 1%.

Luxembourg adopts a repartition model, which means that the currentpensions are paid with the tax incomes from the current workers. Eachgeneration hence pays the pension for the generation before it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 118 / 125

includegraphics[height=0.5cm]lsf.eps

Hypotheses

Salaries grow linearly, S1 with a starting value of 1500 and a growthcoefficient of 3 %, S2 with a starting value of 1000 and a growthcoefficient of 2 %.

Pensions grow also linearly, S1 with a starting value of 2718 and a growthcoefficient of 2%, S2 with a starting value of 1104 and a growthcoefficient of 1%.

Luxembourg adopts a repartition model, which means that the currentpensions are paid with the tax incomes from the current workers. Eachgeneration hence pays the pension for the generation before it.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 118 / 125

includegraphics[height=0.5cm]lsf.eps

Replacement rate in the repartition model

Replacement rate = first pension / last salary

For S1, trep = 27181500(1+0.03)39 ' 57%.

For S2, trep = 11041000(1+0.02)39 = 50%.

A worker who’s trajectory is S1 with a probability of 75 % and S2 with aprobability of 25 % has a replacement rate of

trep =0.75× 2718 + 0.25× 1104

0.75× 1500(1 + 0.03)39 + 0.25× 1000(1 + 0.02)39' 56%.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 119 / 125

includegraphics[height=0.5cm]lsf.eps

Replacement rate in the repartition model

Replacement rate = first pension / last salary

For S1, trep = 27181500(1+0.03)39 ' 57%.

For S2, trep = 11041000(1+0.02)39 = 50%.

A worker who’s trajectory is S1 with a probability of 75 % and S2 with aprobability of 25 % has a replacement rate of

trep =0.75× 2718 + 0.25× 1104

0.75× 1500(1 + 0.03)39 + 0.25× 1000(1 + 0.02)39' 56%.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 119 / 125

includegraphics[height=0.5cm]lsf.eps

Replacement rate in the repartition model

Replacement rate = first pension / last salary

For S1, trep = 27181500(1+0.03)39 ' 57%.

For S2, trep = 11041000(1+0.02)39 = 50%.

A worker who’s trajectory is S1 with a probability of 75 % and S2 with aprobability of 25 % has a replacement rate of

trep =0.75× 2718 + 0.25× 1104

0.75× 1500(1 + 0.03)39 + 0.25× 1000(1 + 0.02)39' 56%.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 119 / 125

includegraphics[height=0.5cm]lsf.eps

Replacement rate in the repartition model

Replacement rate = first pension / last salary

For S1, trep = 27181500(1+0.03)39 ' 57%.

For S2, trep = 11041000(1+0.02)39 = 50%.

A worker who’s trajectory is S1 with a probability of 75 % and S2 with aprobability of 25 % has a replacement rate of

trep =0.75× 2718 + 0.25× 1104

0.75× 1500(1 + 0.03)39 + 0.25× 1000(1 + 0.02)39' 56%.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 119 / 125

includegraphics[height=0.5cm]lsf.eps

Coverage potential in a repartition & capitalization modelWe want to know the sum a that we have to put every year in a savingaccount to get a desired replacement rate taim.

a of course depends on the account’s interest rate i .

If i ∼ U(2%; 7%), a varies between 46 euros and 252 euros with a mean of124 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 120 / 125

includegraphics[height=0.5cm]lsf.eps

Coverage potential in a repartition & capitalization modelWe want to know the sum a that we have to put every year in a savingaccount to get a desired replacement rate taim.

a of course depends on the account’s interest rate i .

If i ∼ U(2%; 7%), a varies between 46 euros and 252 euros with a mean of124 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 120 / 125

includegraphics[height=0.5cm]lsf.eps

Coverage potential in a repartition & capitalization modelWe want to know the sum a that we have to put every year in a savingaccount to get a desired replacement rate taim.

a of course depends on the account’s interest rate i .

If i ∼ U(2%; 7%), a varies between 46 euros and 252 euros with a mean of124 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 120 / 125

includegraphics[height=0.5cm]lsf.eps

Capitalization effort coefficient

τ2 = Sum of the salaries on the salary trajectory / sum of the investmentreturns = 16.5 on average.

That means that you need 16.5 euros from the salary to get 1 euro bycapitalization for the pension.

In fact, if i ∼ U(2%; 7%), τ2 varies between 18 euros and 15 euros.

τ2 =Sj

aj(i − λj)i(1 + i)T − (1 + λj)

T

(1 + i)T − 1.

τ2 depends on a, hence a not only allows to get the desired replacementrate, but a also serves to control the variability of the capitalization effortcoefficient.

We need a compromise between a high replacement rate and a smallcapitalization effort coefficient.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 121 / 125

includegraphics[height=0.5cm]lsf.eps

Capitalization effort coefficient

τ2 = Sum of the salaries on the salary trajectory / sum of the investmentreturns = 16.5 on average.

That means that you need 16.5 euros from the salary to get 1 euro bycapitalization for the pension.

In fact, if i ∼ U(2%; 7%), τ2 varies between 18 euros and 15 euros.

τ2 =Sj

aj(i − λj)i(1 + i)T − (1 + λj)

T

(1 + i)T − 1.

τ2 depends on a, hence a not only allows to get the desired replacementrate, but a also serves to control the variability of the capitalization effortcoefficient.

We need a compromise between a high replacement rate and a smallcapitalization effort coefficient.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 121 / 125

includegraphics[height=0.5cm]lsf.eps

Capitalization effort coefficient

τ2 = Sum of the salaries on the salary trajectory / sum of the investmentreturns = 16.5 on average.

That means that you need 16.5 euros from the salary to get 1 euro bycapitalization for the pension.

In fact, if i ∼ U(2%; 7%), τ2 varies between 18 euros and 15 euros.

τ2 =Sj

aj(i − λj)i(1 + i)T − (1 + λj)

T

(1 + i)T − 1.

τ2 depends on a, hence a not only allows to get the desired replacementrate, but a also serves to control the variability of the capitalization effortcoefficient.

We need a compromise between a high replacement rate and a smallcapitalization effort coefficient.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 121 / 125

includegraphics[height=0.5cm]lsf.eps

Capitalization effort coefficient

τ2 = Sum of the salaries on the salary trajectory / sum of the investmentreturns = 16.5 on average.

That means that you need 16.5 euros from the salary to get 1 euro bycapitalization for the pension.

In fact, if i ∼ U(2%; 7%), τ2 varies between 18 euros and 15 euros.

τ2 =Sj

aj(i − λj)i(1 + i)T − (1 + λj)

T

(1 + i)T − 1.

τ2 depends on a, hence a not only allows to get the desired replacementrate, but a also serves to control the variability of the capitalization effortcoefficient.

We need a compromise between a high replacement rate and a smallcapitalization effort coefficient.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 121 / 125

includegraphics[height=0.5cm]lsf.eps

Capitalization effort coefficient

τ2 = Sum of the salaries on the salary trajectory / sum of the investmentreturns = 16.5 on average.

That means that you need 16.5 euros from the salary to get 1 euro bycapitalization for the pension.

In fact, if i ∼ U(2%; 7%), τ2 varies between 18 euros and 15 euros.

τ2 =Sj

aj(i − λj)i(1 + i)T − (1 + λj)

T

(1 + i)T − 1.

τ2 depends on a, hence a not only allows to get the desired replacementrate, but a also serves to control the variability of the capitalization effortcoefficient.

We need a compromise between a high replacement rate and a smallcapitalization effort coefficient.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 121 / 125

includegraphics[height=0.5cm]lsf.eps

Repartition effort coefficient

τ1 = weighted mean of the salaries on the salary trajectory / weightedmean of the pensions in the repartition model on the pension trajectory =2.7 on average.

That means that the active worker have to earn 2.7 euros to pay 1 euro ofpension by repartition. pause

τ1 =

k(1+d)T+1 PT+1 + ...+ k

(1+d)T+T∗ PT+T∗

S0 + ...+ ST

(1+d)T

.

τ1 depends on the demographic rate d . In fact, if d ∼ U(0%; 5%), τ1varies between 6.7 euros and 1.6 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 122 / 125

includegraphics[height=0.5cm]lsf.eps

Repartition effort coefficient

τ1 = weighted mean of the salaries on the salary trajectory / weightedmean of the pensions in the repartition model on the pension trajectory =2.7 on average.

That means that the active worker have to earn 2.7 euros to pay 1 euro ofpension by repartition. pause

τ1 =

k(1+d)T+1 PT+1 + ...+ k

(1+d)T+T∗ PT+T∗

S0 + ...+ ST

(1+d)T

.

τ1 depends on the demographic rate d .

In fact, if d ∼ U(0%; 5%), τ1varies between 6.7 euros and 1.6 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 122 / 125

includegraphics[height=0.5cm]lsf.eps

Repartition effort coefficient

τ1 = weighted mean of the salaries on the salary trajectory / weightedmean of the pensions in the repartition model on the pension trajectory =2.7 on average.

That means that the active worker have to earn 2.7 euros to pay 1 euro ofpension by repartition. pause

τ1 =

k(1+d)T+1 PT+1 + ...+ k

(1+d)T+T∗ PT+T∗

S0 + ...+ ST

(1+d)T

.

τ1 depends on the demographic rate d . In fact, if d ∼ U(0%; 5%), τ1varies between 6.7 euros and 1.6 euros.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 122 / 125

includegraphics[height=0.5cm]lsf.eps

Systemic risk

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 123 / 125

includegraphics[height=0.5cm]lsf.eps

Global effort coefficient

τ = xτ1 + (1− x)τ2

is the number of euros necessary to pay 1 euro for the pension.

Here x euros come from repartition and 1− x euros from capitalization.

We want to limit the risk of the hybrid system without reducing thepension and in the same time minimize the capitalization effort.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 124 / 125

includegraphics[height=0.5cm]lsf.eps

Global effort coefficient

τ = xτ1 + (1− x)τ2

is the number of euros necessary to pay 1 euro for the pension.

Here x euros come from repartition and 1− x euros from capitalization.

We want to limit the risk of the hybrid system without reducing thepension and in the same time minimize the capitalization effort.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 124 / 125

includegraphics[height=0.5cm]lsf.eps

Aim and solution

Aim : volatility of τ = (volatility of τ1)/m.

Solution:

x =1

m2.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 125 / 125

includegraphics[height=0.5cm]lsf.eps

Aim and solution

Aim : volatility of τ = (volatility of τ1)/m.

Solution:

x =1

m2.

Jang SCHILTZ (University of Luxembourg) joint work with Jean-Daniel GUIGOU (University of Luxembourg) & Bruno LOVAT (University Nancy II) ()Analysis of the salary trajectories in Luxembourg January 19, 2010 125 / 125

top related