Labour market impacts of workfare programmes: Effectiveness of Construyendo Perú

Labour market impacts of workfare programmes
Effectiveness of “Construyendo Perú”
Verónica Escudero
(ILO Research Department and Paris School of Economics)
NEUDC CONFERENCE 2016
November 5-6, 2016
Verónica Escudero “Construyendo Perú” November 5-6, 2016 1 / 18

Motivation
Public works have been increasingly implemented in developing
economies
Peru is a case in point in LAC (19% of all ALMPs, highest share)
Usually in the form of workfare programs
Knowledge from advanced countries cannot necessarily be
extrapolated
High incidence of informality
Inadequate administrative and institutional capacity to implement
policies effectively
Therefore, not enough research exists to unveil the effects of these
policies
Particularly after participation (Jalan and Ravallion, 2003; Ronconi et
al., 2006; Hermani-Limarino, 2011)
Even less is known regarding the effects on job quality
Literature

Contribution of the paper
It measures medium-term effects
– This paper measures the effects in 2012
– Of programme participation during the period 2007-2010
Including the impact on labour market variables
– On employability (e.g. labour market status, type of job)
– But also on work quality (e.g. informality, working time, working
poverty, etc.)
Measure the effects on different groups
– By sex and level of education
Explores heterogeneity in order to identify the drivers of the
observed effects

Key findings
Positive effect on employment (and negative on inactivity)
– Higher employment probabilities of women and the lower-educated
Not so positive effects on work quality
– In some instances, increases in the probability of informality,
excessive work hours and working poverty
Mechanisms
– Resources were not efficiently allocated
– Shift from infrastructure to service-sector projects
– Implementation problems (targeting)

Background

The programme: “Construyendo Per´u”
Objectives:
– Providing temporary employment to the unemployed living in
poverty and extreme poverty
– Improving their employability
Implementation:
– Financing of public investment projects intensive in the use of
unskilled labour
– Public works for a maximum of 4 months
Services provided to participants:
– Temporary employment (685,000 short-term jobs)
– Training opportunities (de facto implemented in 2007 and 2008)
Details on training:
– Soft-skills training was mandatory (yet, not enforced);
– Technical training was voluntary (self-selection)

Targeting: Three stages
Geographical (district level):
– Urban population greater than 2,500 inhabitants
– Ranking of districts according to composite index FAD (factor de
asignaci´on distrital)
– The higher the index the greater the budget assigned to the district
Self-targeting (individual level):
–Wages suﬃciently low
– 352 PEN (252 USD, PPP) for 22 days of full-time work
– 64% of the minimum wage
Individual targeting (individual level):
– Individual characteristics
– Socio-economic situation (thourgh SISFOH)
– Public draw among eligible applicants

Data
1 ENAHO 2007–2013: National Household Survey
– Urban and rural coverage
– 32,000 dwellings and 115,000 individuals per year
– Panel structure for a third of the sample
2 Special survey:
– Carried out in March 2012 to programme participants
– 1142 participants during the period 2007–2010
3 District level database (constructed for this analysis):
– Population, poverty and development indicators;
– Disctricts’ characteristics;
– Disctricts’ year(s) of participation, type of project for which the
district applied and the budget allocated

Speciﬁcation: Fuzzy RDD
Method used:
– Fuzzy RDD given programme assignment
– The running variable is the FAD composite index (which was
reconstructed based on the 3 variables weighted equally)
Control group:
– Non-participants in the vicinity of the discontinuity
Estimators:
– A parametric 2SLS (two-stage least squares)
– A nonparametric LLR (local linear regression) with 3 bandwidths

Speciﬁcation: Discontinuity at the district level
Figure: Mean probability of district participation conditional to the FAD index
(Cut-oﬀ point at FAD=0.125)

Specification: Discontinuity at the individual level
Figure: Mean probability of individuals participating in the programme conditional
to the FAD index (Cut-off point at FAD=0.125)
Identification equation

Graphical eﬀects: Employment and informality
Since RDD is a local estimator, the eﬀect needs to be examined in
the neighbourhood of the discontinuity
Figure: Mean probability of being employed and being employed informally
conditional to the FAD index
working poverty & self-employment

Estimated eﬀects: Labour market status and work quality
All Women
Lower
educated
Higher
educated
Employed
2SLS 2.1 2.3* 4.7* 2.0
LLR 5.1 4.5* 39 4.3
Inactive
2SLS -2.3 -2.5* -4.7* -2.4
LLR -4.2 -4.9* -52 -2.9
Own-account 2SLS 3.6** 2.8*** 1.9 3.5**
worker LLR 5.5** 4.5** 20.7 7.1**
Waged 2SLS -0.03 0.12 0.7 -0.1
worker LLR 3.6** 2.1* 3.7 4.3**
Waged 2SLS -2.8** -1.6** 0.2 -2.8*
employee LLR -7.5*** -3.2* 0.6 -8.9**
Informal
2SLS 5.5** 3.9** 3.3 6.6**
LLR 16*** 7.5** 26 15***
Working 2SLS 7.6*** 5.6*** 0.8 8.9***
poor LLR 14* 10* -21 7.1**
Excessive 2SLS 1.7 1.1 1.3 2.6
hours LLR 14* 11 13 16*

Estimated effects: Interpretation
Clearer effects on women – higher participation of women in the
program
Limited effects on employment of the overall group
Labour market effects might have faded away
There is simply deadweight loss
Detrimental effects on job quality (e.g. informality, working poverty)
Higher skilled – self selection into the technical training, seems to be
promoting self-employment
Women – lack of assistance to raise employability might be
perpetuating unstable labour market patters.
Limited employability – Mostly an income support programme
Sensitivity analysis

Estimated Effects: Hetoregeneity Analysis
Inefficient allocation of resources as a driver of negative effects?
Estimation using programme investment per participant (by
department) as measure of participation
Could changes in programme implementation could account for the
results?
2007-2008: Higher budget and more infrastructre projects
2009-2010: Lower budget and service-sector projects

Baseline
Investment
per participant
2007-2008 2009-2010
Infrastructure
projects
Service-sector
projects
Employed ns ns ns ns ns ns
Inactive ns ns ns ns ns ns
Own-
account
3.6** 4.3** 6.1** 8.3** 5.5** 6.2**
worker (1.5) (1.9) (3.0) (3.2) (2.3) (3.1)
Waged
ns ns ns ns ns ns
worker
Waged -2.8** -3.4** -4.4* -6.0** -3.8* -4.3*
employee (1.4) (1.7) (2.6) (3.0) (2.1) (2.6)
Informal
5.5** 6.7** 9.3* 12.7** 8.4** 9.3*
(2.4) (3.1) (4.9) (5.4) (3.8) (4.9)
Working 7.6*** 9.3*** 12.2** 18.7*** 9.9*** 13.6**
poor (2.7) (3.6) (5.7) (5.6) (3.2) (6.9)
Excessive
ns ns ns ns ns ns
hours
Results for women

Allocation of resources:
Resources do not seem to be financing the most efficient projects
Changes in implementation are likely driving the adverse effects on
job quality of participants
Participants in 2009/2010 (with respect to 2007/2008):
– Higher probability of being employed informally, work as
own-account workers and be working poor
– Lower probability to work as waged employees
Exploring explanations for change in effectiveness between periods:
No evidence of ”frontier effects”
The labour market was not affected by the global crisis in 2009/2010
A change in the nature of investment projects between the two periods
is the most plausible driver of results

Thank you
escudero@ilo.org

Annex 1: Literature
Workfare programs can have an antipoverty effect (Subbarao, 1997)
Direct transfers during participation
Provided wages outweight costs of participation
Stabilization and consumption smoothing effects (O’Keefe, 2005)
Safety nets during crises
Protect individuals from unfavourable decisions even if wages are low
In the longer term, however, effects depend on their employability
effects
If not possible detrimental effects on employment (Hujer, 2004)
Positive effects on poverty if large enough to affect private sector
wages (Dev, 1996)
Motivation

Annex 2: Identification strategy
lim
δ→0
E[Yi | x0 < xi < x0 + δ] − E[Yi | x0 − δ < xi < x0]
E[Di | x0 < xi < x0 + δ] − E[Di | x0 − δ < xi < x0]
= ρ (1)
The causal effect of treatment will be determined by dividing:
– the jump in the outcome-rating relationship
– by the jump in the relationship between treatment status and rating
This will provide:
– an unbiased estimate of LATE
– Wald estimand for fuzzy RD captures the causal effect on compliers
Discontinuity at the district level

Annex 3: Other graphical eﬀects
Figure: Working poverty & self-employment
Graphical eﬀects

Annex 4: Sensitivity analysis
Underlying assumptions:
Continuity assumption and exclusion restriction
The running variable has not been influenced by treatment
The cut-off point has been determined independently
Threats to validity checked:
Agents cannot manipulate the running variable
Cut-off is set independently of running variable
No other discontinuity affecting results
Falsification tests:
Individuals living in rural districts;
Individuals with higher level education;
Individuals in the highest income decile
Evidence of a latent variable driving results?
Differences in district characteristics related to FAD, which did not
change during the duration of the programme
Changes in the budget and institutional factors unlikely to have biased
district participation
Conclusion

Annex 5: Hetoregeneity Analysis
Overall Women
Baseline
Inv. per
participant
2007-2008 2009-2010 Baseline
Inv. per
participant
2007-2008 2009-2010
Employed
ns ns ns ns
2.3* 2.7*
ns
5.1*
(1.2) (1.5) (2.6)
Inactive
ns ns ns ns
-2.5* -2.9*
ns
5.3*
(1.4) (1.7) (2.9)
Own-account 3.6** 4.3** 6.1** 8.3** 2.8*** 3.4** 5.0** 6.2***
worker (1.5) (1.9) (3.0) (3.2) (1.0) (1.3) (2.4) (1.8)
Waged
ns ns ns ns ns ns ns ns
worker
Waged -2.8** -3.4** -4.4* -6.0** -1.6** -1.9*
ns
-3.2*
employee (1.4) (1.7) (2.6) (3.0) (0.8) (0.9) (1.7)
Informal 5.5** 6.7** 9.3* 12.7** 4.0** 4.8** 7.1* 8.7**
(2.4) (3.1) (4.9) (5.4) (1.9) (2.3) (3.9) (3.9)
Working 7.6*** 9.3*** 12.2** 18.7*** 5.6*** 6.8*** 9.4** 13.2***
poor (2.7) (3.6) (5.7) (5.6) (1.9) (2.5) (4.4) (3.8)
Excessive
ns ns ns ns ns ns ns ns
hours
Heterogeneity Analysis

Labour market impacts of workfare programmes: Effectiveness of Construyendo Perú

More Related Content

Similar to Labour market impacts of workfare programmes: Effectiveness of Construyendo Perú

Recently uploaded

Labour market impacts of workfare programmes: Effectiveness of Construyendo Perú