Forecasting of long-term unemployment at the individual level

1
Forecasting of long-term unemployment at the
individual level

• Tomáš Soukup
• RILSA
• Czech Republic
• tomas.soukup_at_rilsa.cz

2
Content

• Introduction
• Theoretical approach (Job Search Theory)
• Data
• Results
• Future developement

3
Introduction

• Macro versus micro approach
• Use of forecasting at the individual level
• defining vulnerable groups at the LM
• forecasting the length of unemployment
• segmentation of claimants
• areas of services at labour offices (zones)
• precautions in social policy
• Predicting the effects of ALMP
• Foreign experience

4
Theoretical approach 1

• Job Search Theory
• H f (V, c U)
• H final result of the job search process,
  matching demand and supply in the labour market
• V number of Vacancies
• U number of Unemployed
• c job search efficiency on the part of the
  claimant

5
Theoretical approach 2

• The success of the job search is principally
  affected by
• The situation in the labour market
• The availability of vacancies and employer
  discrimination (Queuing Theory, Concept of Human
  Capital, Discrimination, Segmentation Theories)
• Job seeker motivation (Nominal Flexibility, Basic
  incentives for work, Concept of Feeling Efficacy)

6
Data and method

• Need for continuous data
• Panel survey 2000 - 2001, 2 waves
• N759
• Binary logistic regression

7
Output variable

• Found a job in 6 months
• 1 found a job
• 0 didnt find a job.
• The probability of finding a job within the next
  6 months was predicted.

8
Main input variables

• Subjective assessment of own chances
• Total duration of past unemployment
• Plans concerning exit from the labour market
• Number of claimants per one vacancy (in region
  and education)

• Nominal flexibility
• Woman with children up to 7 years
• Willingness to change area of work
• Promise of a new job
• School leaver

9
Results 1
Model Nagelkeke R-square Correctly predicted cases
Model with constant only - 54
General model 0.367 73
Model with variables in the PES database 0.192 65
Model with variables in the PES database 5 interview questions 0.331 72
10
Results 2 Prediction versus reality
11
Results 3
Predicted probability of finding a job Job YES Job NO Total N
0-30 13 87 100 266
31-70 52 48 100 303
71-100 83 17 100 189
Total 46 54 100 758
12
Future development

• Analysis of register data
• Prediction of the effects of ALMP
• Scheme for practical use at labour offices

13

• Thank you very much for your attention
• Tomas Soukup
• RILSA, CZ
• tomas.soukup_at_risla.cz