Designing index-based safety nets for village Africa Bart van den Boom Vasco Molini Centre for World Food Studies, VU University Amsterdam Weather Deivatives and Risk January 28, 2010 Humboldt Universität zu Berlin
O utline 1. Introduction 2. Mutual and index-based safety nets for poor villages 3. Semi-parametric index-based safety net for a collective of villages 4. Application to pseudo-villages in Ghana 5. Simulation outcomes 6. Conclusion
1. Introduction Behavioral response to uninsured risks causes poverty traps in village Africa. Hence, the poor are in need of a stronger safety net. The 3 basic safety nets in poor rural communities are : a) Grow low-risk subsistence crops (consumption of home produced staples) b) Obtain an in kind transfer of staple food when own production is too low (consumption from food stocks at village / regional / national level) c) Buy food on the market when own production is too low (consumption from non-farm earnings or from social transfer scheme) Individual and mutual safety net arrangements offer only limited protection, especially in case of a large covariate shock (crop failure; endemic disease). Public institutions are generally too weak to provide an effective safety net for the poor. Costs in avoidance of moral hazard and adverse selection are prohibitive.
1. Introduction (2) As Dercon (2007) observes, the provision of insurance for the poor could well be a key milestone in the fight against poverty But the question is how to do this, given the African setting, with its problems of high transaction cost and weak public institutions. Quest for improving on safety nets in village Africa is urgent, because * Traditional safety nets have eroded (reduced role of chieftaincy) * State interventions have been downscaled (marketing boards; food distribution) * Village food stocks are suited to overcome lean season, but not droughts * Households have few assets, while financial markets are inaccessible to them * Poor population groups like those in Northern Ghana have few safety nets left One idea is to design index-based insurance. The idea is currently receiving attention in the literature and is piloted in various countries, see for example Skees (2008), Molini et al. (2010). In the paper we consider this idea and focus on the design of index-based safety nets, in which insurance is aimed at reducing the risk to fall below a given poverty line. So the aim is not to fully smoothen individual incomes, but to provide social protection for the poor, which will automatically involve (cross-) subsidization.
2. Mutual and index-based safety nets for poor villages Safety nets consist of arrangements that adjust household income, 0()0 τε ψ(,)0 with a premium that is fixed over a range of uncertain events and payments that are contingent on the events that actually happens. 00ϕ 0τ ψ()0 ε Mutual: u(,)h(,)y(,)()y() ε=+ + ε ετεετ Supra-village: 0j 0j 00j 0 The adjusted income becomes:
2. Mutual and index-based safety nets for poor villages (2) The shortfall below the poverty line: l( ε, ε ) = min [ u( ε, ε ) u,0 ] 0 j 0 j Trade-off between expected income and expected shortfall: E u Nº N index N 1 N* full coverage cost N gap transaction cost Safety nets aim at moving from uninsured case towards all-risk insurance: 0 E l N 0 N 1 N gap N index * N
3. Semi-parametric index-based safety net for a collective of villages A safety net for idiosyncratic risks is dealt with by mutual arrangements in the village. An index-based safety net is designed to deal with the remaining covariate risks. The safety net is meant as a low-cost alternative to the all-risk insurance : [ τ*; y * ( ε )] i, where the premium derives from the long-term solvency constraint: τ * solves N max [ u h ( ε ) + τ a,0 ] g( ε ) d ε = τ A + σ i N i i i i i * and this premium suffices to pay for any shortfall: y * ( ε ) = max [ u h ( ε ) + τ * a,0 ] i i i If expected income (plus subsidy) in the collective of villages exceeds the poverty line, then this all-risk premium with its contingent ideal payments would wipe out poverty. However, as noted, this safety net has prohibitive information and transaction cost.
3. Semi-parametric index-based safety net for a collective (2) R() τsubjectony()g()da min y Nmax[uh()y()a,0]g()d εετεε i i ii εεετσ =+ i = + ii i i i F The basis risk that the village faces can be defined as: The all-risk safety net but will not be feasible. [ τ*; y * ( ε )] i would eliminate the entire basis risk, In other words, the admissible set F will not contain the function y*. Following the idea of indexing the payments on a few selected events, denoted x, we propose a flexible semi-parametric function to fit the payments on observed needs.
3. Semi-parametric index-based safety net for a collective (3) The index variables x reflect weather and price conditions. In a more general setting, village and farm specific events could also be allowed for. Unknown parameters f ( x ) = β φ ( x ) + α k( x,x ) j j j s s s Parametric term with given functions φ (e.g. linear, quadratic) Non-parametric term: k is a given kernel function measuring the distance between an event x and the observed events xs
3. Semi-parametric index-based safety net for a collective (4) Minimize the regularized empirical risk: 1 1 R( ξ, η; α, β ) = sξs + λ 2 s rαsαrk( x s,x r ) + ϑη, S The mean absolute error of the regression ys = f(xs) The soft margin with positive scalar theta The quadratic regularization term with positive scalar lambda subject to the regression equation and the self-financing constraint
4. Application to pseudo-villages in Ghana 100 representative villages in 26 states of the world 26 states reflecting economic and weather conditions effectively occurred in the country Main data sources: Ghana Living Standard Survey 1988, 1989, 1992, 1999, 2005 Population Censuses 1970, 1984, 2000 Time series for monthly rainfall data at 40 stations Accra prices for 18 main crops
5. Simulation outcomes Υνινσυρεδ Παραµ ετριχ Poverty rate (%) R-square Std dev ofαμεαν οφ ερρορσ Ρεγ υλαριζ. τερµ Νορτη Σουτη Νορτη Σουτη Νορτη Σουτη Νορτη Σουτη Νορτη Σουτη 0.59 0.42 0.41 0.34 0.55 0.39 122 107 3 0.24 0.23 0.99 0.92 29446 15032 1.2 6.2 0 0 4 0.25 0.23 0.99 0.92 18624 14526 1.3 6.3 43 57 5 0.25 0.23 0.96 0.88 4576 10348 2.7 9.0 174 217 6 0.26 0.24 0.93 0.86 2120 2965 18.8 24.3 750 623 7 0.33 0.28 0.68 0.61 551 408 98.0 78.8 347 286 Βαγγινγ (στδ δεϖ) 0.35 (0.02) 0.27 (0.02) 0.66 (0.03) 0.61 (0.02) 1128 (788) 322 (156) 96.6 (13) 69.1 (9) 205 (20) 290 (51)
5. Simulation outcomes (2) Income distribution for uninsured case, ideal insurance and (semi-parametric) index-based safety net 2 1.5 1.5 0-1 -.5 0.5 1 normalized deviation from the poverty lines Uninsured Semi-Par Param Ideal
6. Conclusion Existing safety nets, mainly mutual at village level, can only provide limited protection. An ideal safety net (an all-risk arrangement against poverty) is infeasible. The low-cost alternative of an index-based safety net may strengthen existing ones. However, the application shows that the basis risk is likely to be large. (moreover, the arrangement requires large cross-subsidization or external funds). The provision of effective safety nets in village Africa is easier said than done.
Three issues 1 Will the scheme provide sufficient protection for poor communities, esp. during crisis?
2. Will the scheme promote the asset base in poor communities?
3. Will the scheme be financially viable? What are the political-economy constraints on (cross-)subsidization between groups? What are the constraints on the level of international solidarity?