Typology of career paths : the insertion period

1
Typology of career paths the insertion period

• Jean-François Giret
• Patrick Rousset

2
Longitudinal career paths the insertion

• Longitudinal career paths and particularly during
  the insertion and post insertion period were
  focused on by a large number of studies at the
  Céreq (centre détudes et de recherche sur les
  qualifications)
• In terms of contingency, calendar
• In terms of structures professional sectors,
  initial training
• In social terms individuals and families
  characteristics
• In terms of policies whether national, regional,
  sectorial
• The study targets defining a typology of youth
  career paths during the insertion period

3
Why a typology of trajectories?

• Focusing on individuals with an holist
  perception
• Considering individuals in their own evolution
• Referring to others.
• Referring to a general context that evolves
  during the period.
• Summarizing and visualizing trajectories
• Linking trajectories to individuals
  characteristics.
• Diploma, gender, social origin, ethnic origin,
  etc.
• Particular Objectives
• Highlighting the descending trajectories (5).
• Taking into account the evolution of fixed term
  contracts from stepping stone function towards
  job insecurity

4
Which hypothesis ?

• Trajectories draw individuals behaviors
  individuals are masters of their own destiny -gt
  holist approach
• Individuals are the input
• The job market context balance between supply
  and demand / between individuals, firms and the
  government
• Trajectories depend on individual rationality
• Ensuing from ones liberties individuals
  rationality makes the structure of the job market
  (demand side)
• The job market structure is deduced from the
  typology output
• Ensuing from ones adaptations to environment
  (supply constraint)
• The job market structure and practices are the
  parameters that influence trajectories and types

5
The Data

• The  Generation 98 survey of the Céreq provides
  a calendar of hiring situations over 7 years for
  16000 young people who left school in 1998. the
  monthly position indicates the status
• 5 status of employment 1 permanent contract PC,
  2 fixed term contract FTC, 3 qualification
  contract, 4 state-subsidized job, 5 temping.
• 4 status of un-employment 6 unemployment, 7
  inactivity, 8 service national, 9 studies.

State distribution plot
Calendar extract
6
The context as an input parameter

• The contracts position individuals on a job
  market but are also vectors of insertion/
  insecurity
• In practice, firms use fixed term
  contracts/temping
• as a trial period mainly with inexperienced
  young people
• as temporary jobs mainly for experienced young
  people
• Evolution in time
• One may benefit from 1 as well as 2 but not at
  the same period of ones life
• In case 1, the Fixed Term Contract mainly
  eventually leads to a permanent Contract unlike
  case 2
• The FTC leads to a PC over the first 3 years and
  declines towards job insecurity
• The FTC is more insecure/secure during
  crisis/economic growth
• State-subsidized jobs vary with government
  policies
• Including context in parameters by
• Establishing a distance between status -gt
  hypothesis that the context is constant in time
• Establishing a distance between status indexed
  with months -gt hypothesis that the context
  evolves during the period

7
Status and state

• The trajectory is encoded with the status
• T(FTC, FTC, FTC, PC, PC, PC) , FTCfixed term
  contract PC permanent contract
• The status refers to the definition
• The state refers to a function
• The situation is the state indexed with time in
  order to consider the evolution of the function
• T(FTC1, FTC2, FTC3, PC4, PC5, PC6)
• Working with situations instead of state is
  necessary to measure the evolve in time of the
  states functions

8
Other methods

• Analyse des correspondances/analyse harmonique
• Euclidean metric c² defined simultaneously the
  distances between situations and trajectories but
  regardless of the short and long term notion (AC)
  
• Correlation between situations defines the main
  paths but the order of the data is not considered
• A H introduces a differentiation long/short term
  but not thoroughly
• Optimal matching
• Matrix of dissimilarity between states prior to
  the trajectories distance
• -gt hypothesis the context is a parameter but
  is constant
• The metric between trajectories adds monthly
  elementary positions independently of paths or
  events (no correlation in time)
• Models of transitions
• Focusing on paths
• The most general exclude individuals basic
  Makovian model (school)
• Allow to measure the evolution of the
  global/individual context
• Define theoretical trajectories
• Allow to consider 1, several, all steps possibly
  with a weighing of the steps

9
Metric between situations

• The situations are the states indexed by the
  months.
• situations are not compared from direct paths
• But from their profile of paths towards future
  situations.
• Weighing the steps the shorter term the
  heavier.
• Distance between the situations is the c2 between
  their profiles of paths towards the future
  (PSS).

10
Structure between situations and evolution of
the job market

• Once the distance between the situations is
  given, one can define the inertia of the
  situations space, the matrix of correlations and
  principal components (Torgerson)
• Principal componentsprincipal events -gt main
  paths
• Matrix of correlations integrates perfectly the
  dynamic of time.

11
Distance between trajectories

• Metric between trajectories is deduced from
  metric between situations.
• Technically
• Projecting trajectories on the principal events.
• Applying Euclidean distance to the new encoded
  trajectories.
• In practice
• All distances are self computed.
• Possibilities of using any Euclidean algorithm
  (clustering Ward, k-means, SOM, linear analysis
  of distances, etc.)

12
The typology with k-means
13
The typology of the careers with SOM

• SOM organizes clusters referring to the proximity
  in the data space on a map
• Units ? graphics displays the states
  distributions plots characterize classes.
• Neighbor classes have similar state
  distributions.
• Proximity shows continuity in time.

A B C D E F G H I
J
1 2 3 4 5 6 7 8 9 10
Cartography of working careers paths
14
Two levels of grouping

• The macro-class grouping (8 clusters) aggregates
  the micro-classes (100 clusters) referring to the
  centroïds distance.

micro-class grouping in macro-classes
Permanent
Towards Permanent
Loss of permanent
Out of permanent
15
Career paths and individuals characteristics
Cartography du diploma
Cartography of gender
Cartography of ethnic origin
Cartography of social origin
Women - High level of diploma - Father
executive Foreign origin
16
Zoom on a path fixed term-gt permanent contract

• Column J shows the timing of the path fixed
  term-gtpermanent contract

A B C D E F G H I J
1 2 3 4 5 6 7 8 9 10
17
C- Path fixed term-gtpermanent
J1 J2 J3 J4
J5 J6 J7
M85 Jan-00 Jun-00 Nov-00 Jul-01 Jan-02 Sep-02 Jun-03
NB 398 288 285 163 173 243 248
Delay . 5 5 8 6 8 9
Flux . 57.6 57 20.4 29 30.4 27.5
M85jan-00
18
Perspectives

• Complementarities with the drifted Markov model
• Two approaches of the context global/individual
  
• Taking into account the context evolution
• Through the same object paths and transitions
• Perspective
• Trajectories initial training insertion
• Discrimination
• Secure/insecure trajectories and working
  environment
• Complete professional careers
• Missing values

19
Typology of career paths the insertion period
20
Description des trajectoiresTransitions et
calendrier

• La déclaration des individus est un récit
  dévènements datés en termes de transitions et de
  périodes de stabilité.
• Dont le calendrier est un recodage arbitraire
• De la définition des 9 états
• Des états hétérogènes CDICDIFonctionnaireindé
  pendant.
• Dont lhétérogénéité varie dans le temps avec
  les proportion et la législation (les contrats
  aidés, SN).
• De la temporalité
• Le temps est découpé en séquences identiques, ici
  le mois.
• La formation initiale est, elle, découpée en
  cycles
• Donne du poids à la stabilité dans les
  classifications
• Dans le but duniformiser la source et les
  résultats
• Contrainte structure adaptée à tous les
  algorithmes.
• Le niveau de nomenclature doit-il être introduit
  en amont du processus danalyse?

21
Données manquantes Censure - Attrition

• Il y a deux stratégies
• La dynamique temporelle de lindividu.
• La comparaison avec les autres.
• Ici, les deux sont réunis
• Projection des situations sur les évènements
  principaux qui, eux, recouvrent toute la période.
• Transitions mesurées à partir des données
  renseignées.

Evolution des corrélations au  CDI en juil2005 
pour b 1/(10.3t) et ? b 1/(1t)