The Evaluation of Chile Solidario

Transcription

1 The Evaluation of Chile Solidario Michael Amior (UCL) Pedro Carneiro (UCL, IFS, CEMMAP,Georgetown) Emanuela Galasso (World Bank) Rita Ginja (Uppsala University) October 2011

2 What is Chile Solidario (CS)? System of social protection for families in extreme poverty (bottom 2.5%) that combines aid and skills development in an integrated approach. (Ministry of Planning) Psycho-social support (24 months) Small conditional cash grant (24 months) Guaranteed cash subsidies Promotion of access to social programmes: social assistance, housing assistance, skills development, employment assistance. Also stimulate the supply of services

3 International enthusiasm with this pioneering approach. Innovative and with potential: direct and integral approach to social exclusion. Inspired adaptations in Colombia, Mexico, Brazil, Peru Lima.

4 But what is the impact of CS: the Design of the Evaluation (WB+MIDEPLAN) Complex program, too many angles (bundle) Best possible non-experimental evaluation design for CS is Regression Discontinuity Eligibility: CAS (or FPS) > Cutoff? Compare individuals just below and above cutoff Cutoff varies with municipality and time Standard difficulties: Sample size? Admin. data large sample. Very local effect? Multiple discontinuities (marginal) Manipulation of CAS score? Really hard to do it Non-standard difficulty: effective ( official) municipality cutoff not observed (Chay et al, 2005)

5 Data: great use of admin. records organized by MIDEPLAN Panel of admin. records: Ficha CAS and FPS Family composition, age and education of each member; geographic location Access to subsidies Employment, Income Assets / Housing characteristics Administrative data from CS Identification of participants Can complement with administrative data on social workers, caseloads, and participation on training/employment programs

6 Eligibility predicts participation

7 Neighborhood Effects? Contagion or congestion (or neither)? Important issue but generally hard to study. Especially important here: RD only looks to marginal individuals. What about the inframarginal? The interaction of heterogeneous preprogram distributions of CAS across communities, and time varying local cutoffs, generates variation in proportions in CS. Can study neighborhood effects mainly for the earlier years of data.

8 Take-Up of SUF (2 years after entry) (use only individuals close to cutoff) (1) (2) CS (t-2) 0.085** 0.152*** (0.040) (0.031) Prop. CS in n (t-1) (0.107) (0.110) SUF *** (0.012) CS*SUF *** (0.021) Prop. CS*SUF *** (0.043) Observations 143, ,043 Standard errors clustered at municipality level

9 Outcomes 2 years after entry, conditioning on pre-program status

10 Impacts of CS years after CS entry Years after entry: Take-up CS 0.189*** 0.124*** 0.231*** of SUF (0.040) (0.044) (0.049) CS*Early *** *** *** (0.023) (0.024) (0.031) Head is CS 0.064* 0.071* employed (0.036) (0.041) (0.044) CS*Early ** ** (0.031) (0.038) (0.038) Legal CS 0.014** 0.032*** 0.074*** occup. (0.006) (0.011) (0.020) of home CS*Early *** *** *** (0.005) (0.007) (0.010)

11 Impacts of CS years after CS entry Years after entry: Employment program CS CS*Fem *** (0.007) (0.010) (0.009) 0.091*** 0.036*** (0.009) (0.009) (0.007) Other interesting points about employment of female heads in target group: - 65% is active in labor market (vs. 95% for male heads) - 20% is cotizando (vs. 33% for male heads) - Cannot detect sustained impact of CS on proportion females active, but - Sustained impact of CS on proportion of female heads cotizando (~6-8%)

12 Social worker quality (fixed effects) is important: SUF Quantile of SW quality Average SUF take-up rate

13 Social worker quality (fixed effects) is important: Head employed Quantile of SW quality Avg. prop. Head emp

14 So what have I learned so far from CS? (complex program, hard to unbundle) Does it increase access to social assistance and employment programs? YES Does it change the employability of the extreme poor? NOT MUCH; BUT FEMALES Remember the target group is: socially excluded, detached from the social welfare system, and unskilled (bottom 2.5%). So, at a minimum I expected an increase in access to social services. But did not expect so much on employment. Social worker effects are large.