This seminar will focus on how countries can establish a policy framework to enable effective local action using an evidence-based approach, choosing between different measures when resources are scarce. The latest evidence from OECD countries on ‘what works’ and ‘what doesn’t’ will be considered, with participants sharing their own experiences from their perspectives as policy makers, researchers, practitioners and social entrepreneurs.
Latest evidence about what works for most disadvantaged - SWEDEN - Christer Gerdes
1. Latest evidence about what works for
most disadvantaged - SWEDEN
Christer Gerdes
Maria Cheung
Swedish Public Employment Service
2. Programme: Job coaching
___________________________________________
• One service in Swedish ALMP is job search assistance,
even called “Job coaching”.
• In 2013 a performance indicator of job coaching providers
was made available at an online website. The indicator,
called “rating”, was based on previous records on how
successful job coaches had been in helping participants
to find work.
• This study examines to what extent this information
affected job seekers choice of job coach.
3. Background
• A job placement officer assess if the unemployed should
have a job coach
• The job placement officer is not allowed to recommend
certain providers of job coaching
• In case the unemployed does not want to choose actively,
she is assigned the job coach closest to her home address
4. The “Rating”
• In April 2013 a performance indicator for providers of job
coaching was published
• The rating is the outcome of comparing the performance of
providers of job coaching during the year of 2012
• The rating is divided in one, two, and three stars
– one star indicate worse than average
– two stars indicate average
– Three stars indicate better than average
5. The “Rating” (cont’d)
• Participants have been “profiled”, which means that by using
a statistical model the chances for each participant of getting
a job were estimated
• The estimated probabilities for participants of getting a job
are compared to the actual outcome, i.e. if participants did
find a job during the time they were assigned to a job coach
• Weighing up these two produce a value (the “score value”)
that is used to compare how well a job coach performed
compared to other job coaches
• Score values are used to rate the coaches in 1, 2, 3 stars
6. 1st January
2012
14th
”Pre-publication” period
November
2012
”Post-publication” period
30th April
2013
Publication
of rating
15th August
2013
Extraction of
data
Rating-period
7. Analysis
• Look at the correlation between participant characteristics
and choice of job coach, controlling for pre-period aspects
Choice of job coach (certain rating value) = X1B1 + X1*[Post-publication]B2 + β[Post-publication]+
X3B3+ε
Variables in vector X1 include indicators for gender, age, education, etc.
Vector X1*[Post-publication] includes same variable interacted with post-publication
indicator
8. 10
Choice of job coach (certain rating value)= X1B1 + X1*[ After publication]B2 + β[After
publication]+ X3B3+ε
VARIABLES Baseline
Female - After publication 0.0756***
(0.0256)
Under 25 years - After publication 0.0346
(0.0472)
50-64 years - After publication -0.00554
(0.0366)
short waiting period - After publication (>41 days) -0.0659
(0.0543)
longer waiting period - After publication (>120 days) -0.00388
(0.0556)
long waiting period - After publication (>256 days) 0.00746
(0.0463)
Very long waiting period - After publication (>632 days) -0.0535
(0.0475)
Compulsory education less than 9 years - After publication -0.0437
(0.0570)
Compulsory schooling 9 years or longer - After publication -0.0290
(0.0390)
Short post-secondary education - After publication 0.0555
(0.0552)
Long post-secondary education - After publication 0.0534
(0.0421)
Born outside Europe - After publication -0.0451
(0.0356)
Control of supply areas Yes
Control for characteristics in pre-period Yes
Number of observations (participants) 12920
R2 0.201
9. Results
• Women seem to have benefited most from publication of
rating values
• Coefficient estimates for education indicators are not
statistically significant, but suggest that better educated
participants choose a job coach with a better rating, and vice
versa
• For participants born in a non-European country there is a
negative correlation (but not statistically significant)
10. Programme: Group meetings
___________________________________________
• Another popular Swedish ALMP instrument is frequent
meetings with caseworkers
group meetings
Individual meetings (face-to-face or digital)
• RCT for “Right Job” implemented at 14 local PES offices
Oct 2011- Feb 2013
Reason: efficiency gains at office level
• This study (Gartell, 2014) evaluates the effects of the
programme wrt
Job-related outcomes
Costumer satisfaction
Caseworkers’ work situation
11. Concept of RJ
___________________________________________
Start seminars
(10-20 participants)
Training on
writing CVs
interviews
Networking etc.
2 w
Start operative teams
(8-12 participants in each team)
4 w 6 w
5 seminars 2 weeks 2 weeks 2 weeks
12. Evaluation technique and method
___________________________________________
• RCT at the individual level
• Encouragement design - ITT estimation
• Treatment: participation in group activities + regular PES activities
• Control: regular PES activities
• Sample:
70 RJ groups were formed at 14 local PES offices
2500 individuals half of which randomized into treatment (~50% take-up)
Aged 18-24 youth registered for ~ 90 days (majority)
30-55 older registered at least 6 months
• Data:
register data
2 surveys: 30 days (78% answer) and 6 months (72 % answer) after
13. Findings
___________________________________________
• Job-related outcomes (survey and register data)
– No effect on job searching but on job proposals
– 3 % higher probability to have a job within 1-3 months but the effect fades out
over time
– Effects stronger for non European born and low educated
– No difference between treatment and control groups after 4 months and onward
• Costumer satisfaction (during treatment, no differences 6 months
after)
– Higher satisfaction levels
– More content with PES service
– More frequent contacts with PES
• Caseworkers’ situation (73 % or 103 answers; 63 worked with RJ)
– Allocated 1.5 more h per unemployed
– No effect on work load or administration burden
– Higher satisfaction levels: content with group activities; better identification of
those with most need; better contact with job seekers
Notes de l'éditeur
The program (“Job coaching”) intends to provide the unemployed an agent whose task it is to help her to find a job, usually by assisting the participant to apply for jobs, to prepare for job interviews etc. The reason for allowing the choice of private provider is to allow for new, innovative, and individually formed ways of providing job counseling activities.
There are unpublished results based on relatively limited voluntary surveys for the Australian Job Network showing that only about a third of consumers used the information in the star rating
The way rating values were introduced does not allow estimating more general causal effects, because there is no possibility to quantity the counterfactual outcome, i.e. what would have happened if the PES would not have published rating values
So, how to measure an effect? For that purpose we use a method called regression discontinuity analysis. The idea behind that approach is to compare those companies that were at the verge of getting one vs. two stars, and three vs. two stars, respectively. We conduct the analysis by studying the closer environment surrounding the threshold values.