The impact of cash transfers on productive activities and labor supply. The case of LEAP program in Ghana

Transcription

1 The impact of cash transfers on productive activities and labor supply. The case of LEAP program in Ghana Silvio Daidone and Benjamin Davis Food and Agriculture Organization of the United Nations Agricultural Development Economics division (ESA) CSAE Conference - Oxford March 19 th, 2013

2 From Protection to Production project Joint project with UNICEF-ESARO Primary funding from DFID ( ) Focus on understanding economic impacts of cash transfer programs in SSA countries Three levels of analysis Table: Country programmes participating in the PtoP project Country Cash Transfer Programme Baseline Follow-up Lesotho Child Grant Program (CGP) Kenya Cash Transfers for Orphans and Vulnerable Children (CT-OVC) , 2011 Ethiopia Tigray Minimum Social Protection Package , 2014 Malawi Social Cash Transfer (SCT) Ghana Livelihood Empowerment Against Poverty (LEAP) Zambia Child Grant Programme , 2013 Monze Cash Transfer Zimbabwe Social Cash Transfer (SCT) , 2015 Daidone, Davis LEAP programme, Ghana 2 / 15

3 Brief description of LEAP Flagship cash transfer programme of Ghana s National Social Policy Strategy. Launched and implemented by the Department of Social Welfare under the Ministry of Employment and Social Welfare Pilot in 2008 with about 1,700 beneficiary households in 21 selected districts. Now scaled-up nationwide with around 65,000 beneficiary households. Districts selected according to poverty maps, communities by a District Implementation Committee Choice of communities based on a range of criteria, including prevalence of adverse health conditions Community based targeting approach for selection of beneficiaries Daidone, Davis LEAP programme, Ghana 3 / 15

4 Objectives of the research Transfers help beneficiary households improve food security and overcome economic constraints associated with missing or poorly functioning markets in credit, insurance and savings Has the introduction of a small but predictable flow of cash income improved livelihood choices and productive investments of vulnerable households? how beneficiaries make decisions regarding the allocation of additional funds impact on agricultural investment in productive enterprises heterogeneity of impact Literature: some evidence in Latin America, not much in Africa. Daidone, Davis LEAP programme, Ghana 4 / 15

5 Analytical framework Objective of an impact evaluation is to attribute an observed impact to the programme intervention Missing data problem bias in estimated impacts deriving from: 1 observable characteristics between beneficiaries and non-beneficiaries 2 unobservable characteristics 3 correlation between unobservable characteristics with both outcome of interest and/or program intervention Daidone, Davis LEAP programme, Ghana 5 / 15

6 Treatment effects of the programme effect for household i: τ i = Y i (1) Y i (0) (1) average treatment effect (ATE) E (τ) = E [Y (1)] E [Y (0)] (2) average treatment effect on the treated (ATT) E (τ D = 1) = E [Y (1) D = 1] E [Y (0) D = 1] (3) Daidone, Davis LEAP programme, Ghana 6 / 15

7 Difference-in-Difference (DiD) estimator For ATE (eq. 2) simple mean comparison is enough in randomized experiments if randomization fails/ observational data and with panel data, ATT (eq. 3) can be improved: E (τ t τ t 1 D = 1) = E [Y (1) t Y (0) t Y (1) t 1 Y (0) t 1 D = 1] = E [Y (1) t Y (1) t 1 D = 1] E [Y (0) t Y (0) t 1 D = 1] (4) Two regression equivalent for the DiD estimator Y it = β 0 + β 1 D it + β 2 R t + β 3 (R t D it ) + ε it (5) ( Yi,t Y i,t 1 ) = Yi = β 0 + β 1 D i + ε i (6) Including baseline characteristics minimize standard errors as long as effects are unrelated to the treatment and constant over time Daidone, Davis LEAP programme, Ghana 7 / 15

8 Inverse Probability Weighting (IPW) When data are affected by covariate unbalancing, Rosenbaum and Rubin (1983) showed that: 1 outcomes are independent of treatment conditional on a vector of observed characteristics (unconfoundness) 2 conditioning on a vector of X is equivalent to conditioning on ˆP(X) IPW works on this idea and consists of using the inverse of ˆP(X) as a weight in the DiD estimator, in order to make treated and control observations representative of the population of interest. This approach is doubly robust Daidone, Davis LEAP programme, Ghana 8 / 15

9 Overview of research design and samples Evaluation strategy for LEAP is a propensity score matching (PSM) design. Baseline data collected from future beneficiaries in three regions (Brong-Ahafo, Central and Volta) Comparison group of matched households (n=699) selected from the nationally representative sample of households surveyed in 2010 by Institute of Statistics and Social Economic Research (ISSER) Additional 215 households were added at follow-up from the ISSER sample Attrition seems not a big deal Table: Samples for LEAP impact evaluation LEAP ISSER Matched ISSER Unmatched Total sample 1,613 1,504 Daidone, Davis LEAP programme, Ghana 9 / 15

10 Propensity score - P(X) We excluded from the probit and from impact analysis units with issues in design (some control households received LEAP!) Figure: Our propensity score Figure: UNC Propensity score regional/community effects in probit estimation NOT included no sensitivity analysis to different types of matching (caliper, nearest neighbor, etc.) and to definition of common support Daidone, Davis LEAP programme, Ghana 10 / 15

11 Balance tests at baseline unmatched LEAP ISSER diff t-stat p-value bias (%) head of hh yrs of education share of sellers # female hhld members over highest yrs of education in hh casual labour cropping and/or livestock farming head of hh widow # orphans living in hhld cropped land - hectares head of hh age matched bias ISSER diff t-stat p-value bias (%) reduction head of hh yrs of education share of sellers # female hhld members over highest yrs of education in hh casual labour cropping and/or livestock farming head of hh widow # orphans living in hhld cropped land - hectares head of hh age The weighting provides a way to adjust the comparison sample Following Hotelling test, we do not reject the null of equal means Standardized differences for all variables except one are below 10. Median bias reduced from 15.9 to 2.4. Daidone, Davis LEAP programme, Ghana 11 / 15

12 Some results -consumption- Daidone, Davis LEAP programme, Ghana 12 / 15

13 Some results -income generating activities- Daidone, Davis LEAP programme, Ghana 13 / 15

14 Some results -financial position- Daidone, Davis LEAP programme, Ghana 14 / 15

15 Conclusions No significant economic impact No impact on consumption No push on agricultural production nor shift towards off-farm businesses or wage labour Constrained households pay down their loans and save Rationale: 1 Social stigma / risk aversion from being indebted 2 Inconsistent implementation of the program 3 Low level of the transfer (7% of per capita consumption) Caveat: study still suffers from limitations! Work in progress... Daidone, Davis LEAP programme, Ghana 15 / 15