unite for children

Can unconditional cash transfers lead to sustainable poverty reduction?

Evidence from two government-led programmes

in Zambia

Sudhanshu Handa, Luisa Natali, David Seidenfeld, Gelson Tembo and Benjamin Davis on behalf of the Zambia CGP Evaluation Team

Presented by Luisa Natali UNICEF Office of Research – Innocenti

September 2016 - University of Florence

2

UCTs in sub-Saharan Africa . . .

UCTs: Regular and predictable transfers designed to support poor people. Paid without any behavioural requirement.

Expansion of Cash Transfers: Worldwide: Almost 800 million people enrolled in

SCTs In SSA: in 2014, 40 countries (out of 50) had some

form of UCT in place up from 21 in 2010

Unprecedented accumulated evidence in SSA

3

Motivation and study aim

What are the broad impacts of UCTs? Unconditional nature of the CT: hhs free to spend money as

they prefer -> impacts may be found virtually anywhere Most studies report results in one specific topical area Given simplicity and coverage of CTs, argument to use CTs

as benchmark for other poverty interventions

Can UCTs go beyond consumption protection and generate productive impacts? (long-term)

4

  The Child Grant Programme   The Multiple Category Targeting Programme

Started 2010   2011

Eligibility requirements

Households with a child under three enrolled

Female- or elderly-headed household keeping orphans; households with a disabled member; other critically vulnerable cases.

Location3 rural districts of Zambia: Kaputa (Northern Province), Kalabo and Shangombo (Western Province)

2 rural districts of Zambia: Serenje (Central Province) and Luwingu (Northern Province)

Cash transferUS$ 12 per month, flat rate, paid bi-monthlyCT around 25% of pre-program consumption

An amount deemed sufficient to purchase one meal a day for everyone in the household for one month.

Two unconditional cash transfer programs in Zambia

Overall goal: reduce extreme poverty and intergenerational transfer of poverty

5

Very similar programs with different demographic profile of households

0.0

2.0

4.0

6.0

8.1

Den

sity

0 20 40 60 80 100Age in years

0.0

2.0

4.0

6.0

8.1

Den

sity

0 20 40 60 80 100Age in years

CGP MCP

preschoolersadolescents

elderlyprime-age adults

6

Zambia implemented ‘gold standard’ evaluations of these two programmes

The Child Grant Programme   The Multiple Category Targeting Programme

Design Longitudinal cluster randomized control trial

Sample size 2,519 households 3,078 households

Unit of randomization CWAC - Community Welfare Assistance Committees (90)

CWAC - Community Welfare Assistance Committees (92)

Method of randomization Public lottery Public lottery

CGP MCP

Baseline 2010 2011

Midline (after 24m)  2012 2013

Endline (after 36m)  2013 2014

7

Methodology and measures• Impacts on around 40 indicators in 8/9 domains of interest at 24 & 36 months

• Multivariate difference-in-differences model • OLS with robust standard errors clustered at the community level• Full household panel

• All indicators defined so that higher values are positive outcomes and converted into z scores relative to the control group

• 2 approaches to account for multiple testing:• for each family of outcomes, we adjust p-values using the Sidak-Bonferroni

adjustment • we build summary indexes as ‘lead indicators’ for each domain

• Robustness checks: ancova, panel fixed effects, Lee bounds• Sample balanced at baseline, no differential attrition

8

DOMAIN Indicators Level

CONSUMPTIONOverall per capita consumption*

HouseholdFood consumptionNon-food consumption

FOOD SECURITY 

Does not or rarely worry about food

Household

Able to eat preferred foodDoes not or rarely eat food he/she does not want to due to lack of resources Does not, or rarely, eat smaller meal than needed Does not, or rarely, eat fewer meals because there is not enough food Never or rarely no food to eat because of lack of resourcesDoes not, or rarely, go to sleep hungry Does not, or rarely, go a whole day/night w/o eatingFood security scale (0-24 where higher means more food secure)*

ASSETS Domestic asset index

HouseholdLivestock indexProductive index

FINANCE and DEBT

Whether woman currently saving cashWomanAmount saved by women

Whether household has new loan

HouseholdReduction in the amount borrowedNot having an outstanding longer-term loan (loans taken out more than 6 months before the follow-up considered)Reduction in the amount owed

INCOME and REVENUES Value of harvest ZMW

HouseholdTotal crop expenditures NFEs [operating or not] NFEs [revenues]

RELATIVE (and/or subjective) POVERTY

Not considering household very poorHousehold

Better off compared to 12 months agoThink life will be better than now in either 1, 3 or 5 years Woman

MATERIAL NEEDSShoes

Child (5-17)BlanketTwo sets of clothesAll needs met*

SCHOOLING School enrolment Child (11-17)Days attended in prior week

NUTRITION OF YOUNG CHILDREN  [CGP only]

Not underweightChild (0-5)Not wasted

Not stunted

9

Results…

10

Effects of CGP on consumption

ITT in actual units

  Food consumption

Non-food consumption

Total consumption1

Impact at 24 months 0.28*** 0.23*** 0.28***

(0.07) (0.07) (0.07)

Impact at 36 months 0.19*** 0.19** 0.20***

(0.05) (0.07) (0.05)

R2 0.25 0.19 0.27

N 6,813 6,813 6,813

Mean standardized ITT

  Food consumption

Non-food consumption

Total consumption1

Impact  at 24 months 0.47*** 0.37*** 0.48***

(0.10) (0.10) (0.10)

Impact at 36 months 0.34*** 0.35*** 0.38***

(0.07) (0.12) (0.07)

R2 0.2 0.17 0.23

N 6,813 6,813 6,813

Notes: Estimations use difference-in-difference modelling. Robust standard errors clustered at the community level are in parentheses. * p<0.1 ** p<0.05; *** p<0.01. Estimations are adjusted and include recipient’s age, education and marital status, household size and household demographic composition, and districts.

11

Total consumption pc

Food security scale (HFIAS)

Overall asset index

Relative poverty index

Incomes & Revenues index

Finance & Debt index

Material needs index (5-17)

Schooling index (11-17)

Anthropometric index (0-59m)

-.2 0 .2 .4 .6 .8Effect size in SDs of the control group

36-month results at a glanceImpacts from both programs similarMCP

CGP

12

Total consumption pc*Food consumption pcNon-food consumption pc

Does not worry about foodDoes not go to sleep hungry at nightDoes not go whole day w/o eating

Domestic asset indexLivestock indexProductive asset index

Harvest value [ZMW]Agricultural input spending [ZMW]Operating a NFERevenues from NFEs

Held any savings (women only)Amount saved [ZMW] (women only) No outstanding debtReduction in amount owedNo new borrowingReduction in amount borrowed

Does not consider hh very poorHh better off compared to 12 months agoLife will be better in the future (women only)

ShoesTwo sets of clothesBlanket

Currently enrolledDays in attendance prior week

Not stuntedNot wastedNot underweight

Consumption

Food security

Assets

Income and Revenues

Finance and Debt

Relative Poverty

Material needs (children 5-17)

Schooling (children 11-17)

Nutrition (Young children 0-59m)

-.2 0 .2 .4 .6 .8 1 1.2 1.4 1.6

CGP (Child Grant) impacts at 36 months

Effect size in SDs of the control group

13

Total consumption pc*Food consumption pcNon-food consumption pc

Does not worry about foodDoes not go to sleep hungry at nightDoes not go whole day w/o eating

Domestic asset indexLivestock indexProductive asset index

Harvest value [ZMW]Agricultural input spending [ZMW]Operating a NFERevenues from NFEs

Held any savings (women only)Amount saved [ZMW] (women only) No outstanding debtReduction in amount owedNo new borrowingReduction in amount borrowed

Does not consider hh very poorHh better off compared to 12 months agoLife will be better in the future (women only)

ShoesTwo sets of clothesBlanket

Currently attendingDays in attendance prior week

Consumption

Food security

Assets

Income and Revenues

Finance and Debt

Relative Poverty

Material needs (children 5-17)

Schooling (children 11-17)

-.2 0 .2 .4 .6 .8 1 1.2 1.4 1.6

MCP impacts at 36 months

Effect size in SDs of the control group

14

  CGP MCP On average

Multiplier effect 1.46 1.72 1.59

Both programmes leverage more money than value of transfer received

Impacts are based on estimated econometric results and averaged across all follow-up surveys. Only statistically significant (at the 5 per cent level) impact estimates are considered. Loan repayments were not measured in the CGP at 24 months.

Household level multiplier =

On average, beneficiary hhs convert each Zambian kwacha of tranfer into an additional 0.59 ZMW worth of income

15

Conclusions• Same overall transformative effects across two programs• Effect not just on primary objective (protective) but also on a range of

productive and economic outcomes -> potential for sustained pathway out of poverty?

• Impacts do not differ significantly across hh eligibility type [external validity]• No impact on anthropometric indicators in CGP• Beneficiaries households spend (59%) more money than value of

transfer itself

• From research to policy: • Impact evaluation evidence played key role in policy changes (scale-up

decisions and budget allocations)• Contributes to cross-country learning under the Transfer Project

16

• Transfer Project website: www.cpc.unc.edu/projects/transfer • Briefs: http://

www.cpc.unc.edu/projects/transfer/publications/briefs

• Facebook: https://www.facebook.com/TransferProject • Twitter: @TransferProjct Email: lnatali@unicef.org

For more information

Ghana, credit: Ivan Griffi