Tajikistan: A Mechanism for Delivering Aid to the Poorest Households The World Bank Subjects of the presentation I. Issue: there is no effective way to distribute aid to the poorest households. II. Background: the national economy, poverty, malnutrition, and migration. III. Evaluation of the present social protection system. IV. Design of the reform. V. Piloting the reform. VI. Evaluation of the pilot. VII.Next steps. I. There is no effective way to distribute aid to the poorest households Motivation: experience during the 2008 food crisis 1. Over the spring and summer of 2008, food commodity prices and especially wheat prices rose sharply. 2. This lowered real purchasing power, public welfare, and put child nutrition at risk in Tajikistan –especially since bread is the staple food . 3. The World Bank and other donors, such as the World Food Program, received funds for use in offsetting the impact of the food price increase on the poorest households. 4. Neither the Government nor donors had any mechanism for identifying the poorest households. 5. There were no lists (no registry) and this blocked donor aid. This led to a request for assistance 1. In 2009, the Government of Tajikistan asked the World Bank for help in building a poverty-targeted social protection system. 2. In response, the World Bank: i. Prepared an evaluation of the current system of social assistance. ii. Prepared a methodology for targeting social assistance iii. Helped the Ministry of Labor and Social Protection to pilot the methodology in two districts. iv. Evaluated the pilot. v. And is funding construction of a National Registry of Social Protection to manage applications, computation of eligibility, and payments. II. Country Background: Geography and its consequences for food security & AID. Tajikistan is a largely mountainous country and only 3 percent of its territory can be cultivated. Tajikistan experiences earthquakes, landslides (especially when the spring melts snow), flooding, and drought –which all harm food security. Tajikistan has a rugged border with Afghanistan, which favors heroin trafficking from Afghanistan through to Russia and beyond. The country could become a target for the Taliban once the US withdraws from Afghanistan. These are the sources of emergencies that harm food security and motivate donors to aid poor households. Position of Tajikistan in Central Asia Economic background: A civil war in the mid-1990s disrupted agriculture and trade and drove down GDP, which has recovered since about 2000. Main industries and sources of poverty COTTON: This was grown in large plantations, which paid low wages to laborers and by small farmers forced to grow cotton. Production is now falling as foreign lenders withdraw financing and because the Government now allows small farmers to decide what crop to grow. Rents accrue to elites. ALUMINUM: Produced a large Soviet-era, electricity-intensive, and government run plant. Rents accrue to elites. HYDRO-ELECTRIC POWER: This is limited by physical capacity. Rents accrue to elites. HEROIN SMUGGLING: Little is known. Rents accrue to elites. SUBSISTENCE AGRICULTURE: Home gardening and herding of sheep and other livestock . Productivity is low. Benefits accrue to the population. CONCLUSION: Large majority of the population derives little income from domestic economic activity. The national poverty rate was 47 percent in 2009 and poverty was concentrated in rural areas and among families with several children 50 49.2 47.2 48 46 44 41.8 42 40 38 Tajikistan Urban Rural Poverty rates in 2009 70 62.4 60 50.9 50 36.5 40 30 29.5 20 10 0 None One Two Three Household poverty rate by number of children. Tajikistan: nutrition situation Underweight (low weight/age) in children aged 0 to 59 months was 12.5% in July 2011 (WFP). Wasting (low weight/height) is a problem particularly for children less than 24 months of age (WFP). Stunting (low height/age) was 29 percent among children aged 6 to 59 months (UNICEF and the Ministry of Health, 2009). Note: The high stunting rate means that infants and young children suffer from chronic under-nutrition, which severely and often irreversibly affects not only the physical but also cognitive development of children. Stunting in the first two years of life is especially damaging as it increases children’s vulnerability to disease and early death. If the children survive, they face poor physical health throughout their lives and limited capacity to learn and earn an income. Main survival strategy: migration Poverty and malnutrition spurred a mass migration of men to Russia. A large fraction of the population are, in effect, economic refugees who have left Tajikistan to earn a living for themselves and their families. The exact volume of migration is not known, but the World Bank Office in Tajikistan estimates that about 1.5 million people, mostly men, out of a total population of 6.8 million people have migrated to Russia to work. Migrants sent home about USD 2.3 billion in 2010, which contributed greatly to the survival of the population. III. Evaluation of Tajikistan’s Social Protection System: Tajikistan spends less on social assistance, as a percentage of GDP, than any other country in the region. , Hungary 06 Croatia 08 Bosnia 08 OECD 05 Belarus 08 Uzbekistan 07 Ukraine 08 Serbia 08 Russia 06 Albania 08 Romania 07 Estonia 06 Lithuania 08 Kosovo 08 Poland 07 Moldova 08 Kazakhstan 07 Bulgaria 08 Kyrgyzstan 08 Georgia 07 Montenegro 08 Armenia 08 Macedonia 08 Latvia 08 Azerbaijan 08 Turkey 07 Tajikistan 09 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% Share of social assistance spending in GDP in ECA Countries (in percentage points) 4.0% The main social assistance program compensates households for increases in electricity and natural gas prices. Table 1: Budget expenditures on social assistance in 2009 Program Annual budget in US$ millions Share in GDP, in percentage points Social Assistance $9.96 0.20% $22.18 0.45% $12.22 0.25% $4.87 0.10% $2.86 0.06% Social Assistance plus Social Pensions Social Pensions Main social assistance programs: Electricity & Gas Compensation Compensation to needy families whose children study in school (Conditional Cash Payments) Sources: Ministry of Finance, State Agency for Social Insurance & Pensions, World Bank Social assistance benefits are a tiny proportion of the income of the poorest quintile of households 60 56.8 52.0 50 45.8 45.0 43.4 39.8 40 30 36.1 35.4 33.6 31.6 30.1 28.9 27.1 26.6 26.3 25.4 24.3 23.5 20.6 20 11.2 10.1 10 9.4 2.9 0 Source: World Bank, ECA data base Generosity - Social assistance as a share of the income of the poorest quintile in ECA counties (Percentage point shares) Coverage rate: Social assistance benefits reached only about 20 percent the poorest quintile (20 percent) of households. The rest leaked to higher better off groups. 100 90 80 70 94.4 86.3 75.7 72.7 69.5 69.0 67.2 60 50 57.9 55.2 54.3 54.2 53.3 52.4 50.1 44.3 44.3 40 30 20 10 0 Source: World Bank, ECA data base Coverage of the poorest quintile of households. 39.2 36.9 31.9 31.7 31.0 28.3 22.0 21.0 19.9 The case for reform is overwhelming. The consolidated budget shows that social assistance is relatively small. Furthermore, analysis of the 2009 data shows that the current social assistance programs are relatively ineffective in reaching the poor and in reducing poverty. And finally, the review of implementation of the programs identified slips in budgeting, targeting, and accountability. IV. Design of the reform Overview of the reform The Government decided to pursue the following strategy: Consolidate the present programs into a single social assistance program. Adopt a system of targeting of social assistance. Manage applications, decisions on eligibility, and payments through an National Electronic Registry. Proxy-means test for targeting. Benefits could be targeted by region, by using and income test, a proxy-means test, or mixed-tests. The Government rejected geographic targeting of the poorest districts on the grounds that this was inequitable. In any case, no poverty data are available at the district level. Income tests are not practical because households will not report receipt of remittances from abroad and it’s difficult to estimate household income from subsistence agriculture. So the Government agreed to apply a proxy-means test (PMT). The World Bank team estimated the test from household budget expenditure and asset data. Variables used in the best performing models Characteristic Household size Gas oven Generator Electric radiator Refrigerator Satellite dish Car or truck Computer Household Head's Employment Sector Household Head’s Education Housing Roof Material Number of Children under 15 Oblast Total Number of Variables Used for Urban Model * * * * * * * * * * * 11 Source: World Bank analysis of the Tajikistan Living Standard Survey of 2009. Used for Rural Model * * * * * * * * * 9 Variables and Weights for the proxY-means tests Urban Variable Log household size Electric radiator Refrigerator Computer Satellite dish Car or truck Number of children under 15 years Employment in Agriculture, Fishing or Forestry Employment in Manufacturing or Mining Employment in Services (utilities) Employment in Construction Employment in Public Admin, Health or Educ. Employment in Sales and Services Weight -0.5694 0.2333 0.2135 0.2354 0.2399 0.3137 -0.0253 0.2389 0.0427 0.1548 -0.0109 -0.0215 0.1141 Sector of Employment = Other Household head’s education = basic Household head’s education = secondary Household head’s education = higher Housing roof material = metal sheeting, tiles, mud, concrete slab Housing roof material = thatch oblast = Sughd oblast = Khatlon oblast = RRP oblast = GBAO Constant -0.0375 -0.1727 -0.0107 0.1081 -0.0846 0.2684 0.0038 0.0526 0.2395 -0.0616 5.8783 Source: World Bank analysis of the Tajikistan Living Standard Survey of 2009 Rural Variable Household size = 4 to 5 Household size = 6 to 7 Household size = 8 to 12 Household size = 13 or more Gas oven Generator Electric radiator Car or truck Satellite dish children under 15 = 1 or 2 children under 15 = 3 children under 15 = 4 to 6 children under 15 = 7 or more Weight -0.2182 -0.3063 -0.4412 -0.6043 0.0801 0.2033 0.3441 0.2203 0.1915 -0.1041 -0.2329 -0.3391 -0.4688 Housing roof material = metal sheeting, tiles, mud, concrete slab Housing roof material = thatch oblast = Sughd oblast = Khatlon oblast = GBAO Constant -0.1865 -0.0249 -0.1258 -0.1448 -0.2360 5.6923 Performance of the proxy-means test model in simulation Settlement Type N R2 Predicted error of Inclusion Exclusion Predicted targeting accuracy Urban 539 48.8% 42.3% 31.0% 57.7% Rural 963 29.2% 47.5% 50.4% 52.5% All 1,502 . 45.9% 45.4% 54.1% Note: The scoring formulas are estimated for rural and urban areas separately. The target group is the lowest quintile of per-capita consumption. Source: World Bank analysis of the TLSS 2009. ‘N’ is the number of households in the sample. ‘R2’ is the percent of variation in consumption per capita that’s explained by the formula (estimated regression equation). V. Pilot of the reform Proxy means-test for targeting. The Government launched a pilot program of consolidated and poverty-targeted social assistance in Yevon and Istaravshan districts in 2011. It targeted the nationally poorest 20 percent of households using the proxy-means test (above). Between January 2011 and January 2012, 18,299 applicant households applied for the pilot social assistance benefit in the two districts, representing approximately 116,000 household members. The Government paid social assistance benefits to the households that the proxy-means test identified as poor. VI. Evaluation of the Pilot: The evaluation draws on two large surveys In the two pilot districts, enumerators interviewed 2,000 applicant households and 500 non-applicant households. In a national survey, enumerators interviewed 2,000 households. Interviews obtained detailed information on consumption spending, assets (including livestock), housing characteristics, education, and health. Estimated results of the pilot National Quintile Group 1 2 3 4 5 Total Population in Istaravshan and Yevon 162,338 81,896 60,864 61,032 37,310 403,440 Share of population in each quintile 40% 20% 15% 15% 9% 100% Probability someone knows about the program in each quintile 54% 42% 47% 37% 42% 47% Probability someone applied in each quintile 30% 25% 25% 25% 42% 29% No. of applicants (incl. household members) 49,105 20,760 15,285 15,112 15,630 115,891 73% 73% 62% 54% 37% 64% 35,991 15,081 9,511 8,235 5,851 74,669 Coverage; probability someone was beneficiary in each quintile 22% 18% 16% 13% 16% 19% Targeting: share of benefits going to each quintile 48% 20% 13% 11% 8% 100% Probability of application being selected No. of selected beneficiaries (incl. household members) Coverage of the poorest quintile is low because too few of them applied 30 percent of the poorest quintile applied. Only 54 percent of the poorest quintile knew about the pilot program. Nevertheless, the coverage rate for the target group (the nationally poorest 20%) was still higher than in the existing programs. Less than 30% of applicants in lowest quintile (the poorest); more than 40% applied in the highest quintile Proxy-Means Test performed relatively well The PMT formula performed well in identifying 48% of the nationally poorest 20% in the pilot districts. Most of the inclusion errors are in the second quintile. The PMT performance was particularly high when considering those who were among the poorest 40% nationally: 48% of the poorest applicants enrolled (23% in current programs). Proxy-means test: Who was selected? The percentage of accepted applications is highest at the lowest quintile, and lowest for the highest quintile Main conclusions from the evaluation. The proxy-means test formula performed better in identifying the eligible beneficiaries than the current programs on conditional cash transfers, electricity and gas compensation. Coverage of the poor is unacceptably low. Ways to improve targeting of social assistance Improve the proxy-means test through reestimation from 2012 data. Mobilize and train local officials and community leaders to disseminate information and collect applications. New Proxy-means test 2012 National Survey could lead to a more accurate test because it includes new variables Housing construction and components. Plumbing and water supply. More detail on farm animals. Cows, bulls, male goats, female goats, donkeys, rabbit, beehives, poultry, fish for aquaculture, sheep, pigs. Variables most correlated with consumption Both Urban and Rural • Landline phone • Generator • Electric radiator • Fridge • Washing machine • Electric heater • Microwave • Electric sewing • Computer • Satellite dish • Car or truck • Chimney • Number of people in HH • Living density Urban Only • Electric oven • Electric stove • Metal bars window • HHH education • Labor force status • Children under 11 years • Housing Floor • Marital Status Rural Only • Gas cylinders • Gas oven • Broshooka • Electric Fan • Housing Foundation • HHH age • Children under 14 Performance of 2009 vs 2012 0.7 PMT Coverage Rates by Quintile 0.6 0.5 0.4 2012 0.3 2009 0.2 0.1 0 Q1 Q2 Q3 Quintile Q4 Q5 Community outreach to increase the number of applications from poor households The information campaign should target the poor community through training of local civil servants and leaders, rather than target the broad public through TV, radio, and newspaper ads. Mobilize traditional community leaders (mahalya rais), women’s committees, and other community leaders to: Disseminate information on the pilot. Gather applications. Check the accuracy of information (to avoid rejection of applications at the district level). Measuring impact of social assistance reform in Tajikistan using regression discontinuity 18,298 applicants, representing 115,891 people, had applied in the pilot districts by March 2012 This is 29% of the population of the pilot districts Transfer is 100 Somoni Per Quarter (21 USD) 7 USD per month per household Represents 2.4 % of average household consumption in the pilot districts Evaluation Methodology: Regression Discontinuity Compare the outcomes (e.g. food consumption) of eligible and non-eligible applicants…. around the 180 threshold PMT score Two detailed surveys in 2012: Pilot district survey: interviewed 2000 applicant households and 500 randomly selected nonapplicant households National survey: interviewed 2000 randomly selected households Preliminary results from regression discontinuity tests Strong results Food consumptions increased. No significant impact on employment, or education and health expenditures Weak Child labor decreased in female-headed households. School absenteeism decreased in female-headed households. These results would be surprising, if confirmed, considering the small size of the social assistance benefits. VII. Next Steps The Government plans to expand the piloting to 8 additional districts. The expanded pilot will apply the proxy-means test estimated from 2012 data, and adjustment to include more elderly households. There will be an outreach program through training of civil servants and use of community leaders to spread information and gather applications. Construction of an IT system for management of applications, eligibility decisions, and for financial controls on payments