Caterina Ruggeri Laderchi, Ramya Sundaram, Natsuko Kiso and Alexandru Cojocaru World Bank International Conference “Poverty and Social Inclusion in the Western Balkans” Brussels, Belgium, December 2010
Better household level information can improve performance of targeted programs at no cost Criteria anchored in a poverty measure can be useful even when sudden change is not feasible Clear and objective criteria improve transparency and support for a program
Social assistance in Albania (Ndihma Ekonomike) Current targeting mechanisms Targeting performance Results of simulations ◦ Block grant (geographic) allocations ◦ Household level identification Conclusions
Largest non-contributory social assistance (cash benefit) program in Albania ◦ Over 100,000 HH in 2008 (7% of population) ◦ BUT budget of only 0.3% of GDP Administered by local governments ◦ Block transfer from central government ◦ Centrally defined identification rules ◦ Local approval of eligibility and distribution of benefits
CriteriaNE allocation scheme Block grants ◦ Regional poverty estimates (LSMS 2008) and municipal population estimates (Census 2001) ◦ # of NE beneficiaries in municipalities in previous year Household identification ◦ Means-test, implemented through multi-layered filters ◦ Different across urban / rural areas NE budget Communes Households Block grants Household identification
Relative to neighboursImproving over time
Coverage of bottom quintileCoverage over time
Block allocations per poor person Urban program: amount of transfer is fixed Urban poverty rates are now high in coastal areas But allocations are not adjusted accordingly
Two counterfactual simulations ◦ Geographic targeting: through a poverty map ◦ Household level targeting: through a PMT Main features of the counterfactuals : ◦ We focus on one feature at the time ◦ We simplify by using Per capita allocations (no equivalence scales) No differences between rural/urban amounts
CounterfactualSimulation Geographic allocation simulation Geographic targeting Proportional to actual weights Proportional to poverty index Household targeting Current NE recipients Household level targeting simulation Geographic targeting No geographical allocation: total budget/number of beneficiaries Household targeting Current NE recipientsBottom 7 percent of population as identified by PMT
Targeting accuracy with actual weights Targeting accuracy with poverty headcount weights Targeting accuracy with poverty severity weights Q Q Q Q Q Poor Non-poor
Replacing filters with household eligibility based on a proxy means test (PMT) ◦ Household composition ◦ Type of dwelling ◦ Asset ownership Identify the (predicted) bottom 7% of population Allocate within current NE budget envelope
Simulation resultsAssumptions Coverage of poorest decile (%) Share of benefits to poorest decile (%) Baseline PMT No geographic targeting ◦ Benefit = budget / # beneficiaries Constant overall budget Constant overall share of beneficiaries
Advantage of improving geographic targeting with the poverty map ◦ Improved targeting ◦ Improved transparency – even if not jumping to a new system ◦ Further improvements likely with new Census data Additional improvement possible in the long run with a centralized national criterion for identifying beneficiaries