How Much? Spending on SSN Programs Cem Mete Senior Economist World Bank December 6, 2011 1
Outline 1. The macro decisions: how much to spend on safety nets? 2. At the program level: how much to pay? Benefit level considerations by type of program; benefit formula; whom to pay?; inflation and program generosity 3. Paying too little --- is it possible to get (some) results? Impact evaluation of a pilot CCT program. 4. Paying too much --- (when) should one worry about work disincentives? 2
How much to spend on safety nets? Start with a strategy that actually prioritizes among different objectives Is it all citizens have a right to type of thing? Is it about addressing the basic needs of chronic poor? Is it about assisting the transitory poor deal with shocks? Is it about increasing social mobility through investing in next generation s schooling and health? 3
How much to spend on safety nets? Considerations for the overall budget envelope include: Level and structure of poverty Tax base (informal sector ) Underlying distribution of productive ability (exiting poverty ) Institutions for private provision Quality of public service delivery The nature of shocks affecting the country Politics 4
In practice, there is a lot of variation in percent of GDP spent on safety nets, by GDP per capita (ppp) and dependency ratio (2008) 5
Grosh et al (2008) do highlight some correlates of safety nets spending, with no statistically significant linkages (although there are stronger trends for social protection and social sector spending) noting that mean spending on safety nets is 1.9 percent of GDP 6 Factor Per capita GDP (PPP) Safety Net Spending as % of GDP Social protection spending as % of GDP Social sector Spending as % of GDP 0.0768 0.5045** 0.5460** Gini coefficient -0.1104-0.3410** -0.2686* Voice 0.0678 0.2294** 0.2607** Ethnic fragmentation 0.1628-0.0204-0.0972 Democracy 0.1733-0.0533 0.1907 Attitudes about inequality 0.1234-0.1694-0.1559
At the program level: How much to pay? Basic question No simple answer: Result of an iterative process of program design 7 Benefit level: maximize desired program outcomes: smallest transfer needed to achieve the desired improvement in the targeted outcomes Function of available budget, administrative capacity political constraints Hence setting the benefits will be program specific
Cash transfers Type of program Guaranteed minimum income Last resort programs Food stamps Family allowances Heating allowances Social pension Benefit level depends on: Eligibility threshold income of beneficiary household Poverty gap/ Cost of an adequate food basket Food poverty gap The cost of raising a child Seasonal increase in the heating cost during cold season Poverty line Minimum contributory pension 8
Conditional cash transfers Type of grant Education grant or scholarship Benefit level depends on: Opportunity cost of the time spent by the child in school (child labor earnings) Direct costs of schooling Health and nutrition grant Opportunity cost of the time spent by mothers on health checks / nutritional education 9 Supply incentive Cost of improved service (wages, material costs)
Workfare Wage level General case depends on Below wage level for unskilled workers Often, the number of days provided by an individual worker are rationed If higher than wage of unskilled labor, need add l targeting mechanism to ration the demand 10
In-kind transfers Type of program Benefit will depend on School feeding programs Cost of the food bundle + Logistical costs Food ration rationale: to reduce the food poverty gap of beneficiaries Same rationale as for lastresort programs + logistical costs 11
Benefit level in practice Result of a trade-off Given budget, determine benefit level and program coverage Benefit level should be: neither too high to generate dependency, nor too low to lack impact 16% of the population is poor Reconsider benefit level and coverage The SSN budget envelope is only 1% of GDP Their average consumption is 25% below the poverty line To cover this consumption gap I need 4% of the GDP 12
Benefit levels in practice Comparisons are difficult 13 Comparative evidence is scarce Comparison across programs and countries is difficult. Such information is presented as: Level of benefits expressed in local currency, when variable formulae presented at a table Level of benefits in comparable purchasing power (USD PPP) But generosity is a relative concept, differs from country to country In relative terms: % of min wage, average wage, poverty line, unemployment benefit, social pension Our preferred measure: Generosity = benefit / consumption of beneficiary household
Benefit levels in practice Limited to 4 types of programs, 2 regions Family allowances (n = 15) Last-resort programs (n = 20) Conditional cash programs (n = 6) Social pensions (n = 14) 14 Generosity of Social Safety Net Programs from ECA and LAC Regions 0 20 40 60 Benefit in % of the consumption of recipient household Source: Tesliuc (2008, forthcoming) and Shady (2007)
Benefit formula 15 Benefit levels may vary by: The poverty level of the family / household Family size, composition Age of the family members Gender Over time / Seasonal Higher in the lean or hungry season Higher in the cold season (heating costs) Region Time spent in the program
Factors taken into account in the benefit formula, CCT programs 16 Benefit varies by HH structure Duration Payee Frequency of HH # of cap other HH Age of in payments Country/Program income children members children Gender program Kenya CT for OVC max=3 bimonthly Cambodia JFPR parent/guardian quarterly Turkey SRMP mothers bimonthly Brazil Bolsa Familia max=3 mother monthly Chile Solidario head of household monthly Colombia Familias en Accion mother bimonthly Dominican Republic Solidaridad head of household bimonthly Ecuador BDH women monthly Honduras PRAF II mother quarterly Jamaica PATH family representative bimonthly Mexico Oportunidades mother bimonthly West Bank Gaza Bangladesh FSSAP female student twice a year
Paying too little --- is it still possible to get (some) results? Pakistan conditional cash transfer (cct) pilot --- illustrating the tradeoff between increased coverage versus more assistance to the poorest Implemented by Pakistan Bait-ul-Mal (PBM), as Child Support Program (CSP) Poor households with children between the ages of 5 to12 years old are targeted Initially piloted in two (remote and poor) districts, later expanded to eight more with some design improvements Education conditions for the CSP require that beneficiary children between the ages of 5 to 12 years must be enrolled in school and maintain regular (80%+) attendance. 17
Lots of problems during first phase implementation, including Very small benefit amount (varying from 300 to 500 rupees per quarter, in addition to the food support program cash transfer of 3000 Rupees per year) Serious delays in delivering payments to beneficiaries due to documents required to process payments, late receipt of attendance sheets from teachers etc. Lack of a follow-up (utilizing social workers etc) with eligible households that don t comply with the conditions. Limited partnerships with provincial education departments to ensure that school supply conditions are adequate. 18
Difference in difference primary school enrollments (with Xiaohui hou, Kinnon Scott and stefania cnobloch) 19
20
Selected Findings The CCT pilot led to 11.65 percentage points increase in enrollments despite very small benefit amount but (as discussed below) it seems the benefit amount that did not fully compensate for the value of children s time was not sufficient to increase enrollments in poorest households that rely on their children s employment. Even though the program did not distinguish between males and females, in practice it ended up increasing female enrollment rates much more than male enrollment rates: 13.74 percentage points versus 9.06 percentage points The CSP led children from the poorest 40 percent households increased school enrollment rates by 12.03 percentage points, while for children from wealthiest 60 percent the effect was slightly smaller at 10.85 percentage points. Here, it is important to point out that in 2006 (at the time of the baseline survey) children from the wealthiest 60 percent of the population had higher enrollment rates in both control and treatment districts to start with The CSP seems to have had no impact on the employment patterns of children between 9 years old and 12 years old (not shown) 21
Inflation and program generosity 22 Issue for cash programs (incl. CCT, workfare) During crises Benefits of SSN programs are, in general, indexed less often than pensions, unemployment benefits Good practice: Automatic indexation Revision of benefits = f(domestic inflation)
Whom to pay? Head of household most common Any applicant from a household 2 nd most common New evidence suggest that the choice of the beneficiary within the household matters Women will spend more on child welfare and are less likely to favor boys over girls Girls are more likely to attend secondary school if cash go to them rather than their mothers 23
Paying too much: (when) should one worry about work disincentives? 24 According to theory, any unearned income lowers incentive to work via the income effect. Some programs, that reduce the level of benefits when the earnings go up, also lower incentive to work via the price effect. The price effect is at its theoretical maximum in a simple guaranteed minimum income program: Everyone who earns less than the guarantee gets a transfer just large enough to fill it; Implies 100% marginal tax rate for recipients, and that their work effort will fall to zero
Concern over work disincentives may be overstated This argument is politically very powerful, but actually overstated 25 Its largely a developed country concern and literature: Sizable work disincentives were found in a program for lone parents (mother and children) from US, Aid for Families with Dependent Children: 10-50% reduction in work effort Design features encouraging dependency in AFDC: It had 100% marginal tax rate Was generous (income-replacement program) Its target population was single mothers, who for most of program history society wanted to support in their child raising role. Since then, 1996 reform to TANF program and EITC provide many work incentives 25
Concern over work disincentives may be a little overstated Large work reductions are not generalized to other programs in developed countries US constitutes bulk of evidence Most studies of food stamps, nutrition programs, child care subsidies, housing, Medicaid find no evidence of reduced work efforts; 26
Why we think that the concern over work disincentives may be overstated for developing countries? Very few programs in developing countries operate as simple GMIs; The few GMIs often have workfare or activation requirements; Few programs are means tested; Benefits are rarely more than 20 percent of base welfare so survival on them alone is impossible Programs often limited to those not meant to work: children, elderly, disabled 27 27
Evidence from developing countries Is limited: CCT: Progresa, Mexico (Parker and Skoufias, 2000; Skoufias and di Maro, 2006; Freije, Bando, Arce, 2006) no sig. impact CCT: Bolsa Familia, Brazil (Leite, 2006) no sig. impact In-kind & cash transfer: Food Support Program (PAL), Mexico (Skoufias, Unar, Gonzalez-Cossio, 2008) no sig. impact Food aid, Ethiopia (Abdulai, Barrett, Hoddinott, 2005) no sig. impact Rice Subsidy Program, Sri Lanka, (Alderman and Sahn, 1996): 10% reduction in adult labor Synthesis of effects of suite of CCT programs (Fiszbein and Shady eds, forthcoming) finds no evidence of reduced work effort by adults sometimes but not always meaningful reductions in child labor 28
29 Thank you!