Donor coordination is a problem in social choice. There are several donors, each of whom wants to donate some money. Each donor supports a different set of targets. The goal is to distribute the total donated amount among the various targets in a way that respects the donors' preferences.
As an example, consider a town with three recreational facilities that require funding: theater, chess club, and basketball field. There are two donors: Alice and George, each of whom wants to donate 3000. Alice wants to donate to indoor activities (theater or chess), whereas George prefers to donate to competitive activities (chess or basketball). Suppose further that the donors consider the facilities substitute goods, so that the utility of a donor is the sum of money distributed to a facility he likes. Consider the following possible distributions:
Alternatively, one can assume that the donors consider the facilities complementary goods, so that the utility of a donor is the minimum amount of money distributed to a facility he likes. In this case, the uncoordinated distribution 1500,3000,1500 gives both donors utility 1500; the distribution 0,6000,0 gives both donors utility 0; but there is an even better distribution:
In both cases, coordination can improve the efficiency of the allocation.
Donor coordination is a variant of participatory budgeting, in which the budget is donated by the voters themselves, rather than given by the government. Since the donations are voluntary, it is important that the coordination algorithm ensures that each voter weakly gains from participating in the algorithm, i.e., the amount contributed to projects he approves of is weakly higher when he participates than when he does not.
Donor coordination has been studied in several settings, which can be broadly categorized into divisible and indivisible:
Donor coordination with divisible targets is similar to the problem of fractional social choice, except that in the latter, the "budget" is fixed in advance (e.g. time, probability, or government funds), and not donated voluntarily by the agents.
Brandl, Brandt, Peters and Stricker[1] study donor coordination with additive binary (dichotomous) preferences, represented by approval ballots. Formally, for each donor i there is a set of approved charities denoted by Ai, and i
ui(d)=
| \sum | |
| x\inAi |
dx
They analyze several rules. They are exemplified below for a setting with 4 targets (a,b,c,d) and 5 donors who contribute 1 each, and whose approval sets are ac, ad, bc, bd, a.
They also prove a strong impossibility result: there is no PB rule that satisfies the following three properties: strategyproofness, efficiency, and positivity (- at least one approved project of each agent receives a positive amount). The proof reasons about 386 preference profiles and was obtained with the help of a SAT solver.
Brandl, Brandt, Greger, Peters, Stricker, Suksompong[2] study donor coordination assuming donors have additive but non-binary utilities. Formally, for each donor i and charity x, there is a value vi,x, and i
ui(d)=
| \sum | |
| x\inAi |
vi,xdx
They prove that the Nash product rule incentivizes donors to contribute their entire budget, even when attractive outside options are available. while spending each donor’s contribution only on projects the donor finds acceptable. The Nash rule is also efficient. On the down side, it is not strategyproof, and violates simple monotonicity conditions (even in the binary case).
Brandt, Greger, Segal-Halevi, Suksompong[3] study donor coordination assuming donors have Leontief utilities. This is motivated by funding charities, where it is reasonable that donors want to maximize the minimum amount given to a charity they approve. More generally, for each donor i and charity j, there is a value vi,j, and i
ui(d)=
| min | |
| x\inAi |
dx/vi,x
They define a rule called the Equilibrium Distribution Rule (EDR), which finds a pure-strategy Nash equilibrium in a game in which the donors' strategies are the possible decompositions of their donations. They prove that there always exists a unique pure Nash equilibrium, and it can be found efficiently using convex programming, by maximizing the Nash social welfare (a sum of logarithms of agents' utilities, weighted by their donations). EDR is Pareto-efficient, group-strategyproof, and satisfies several other monotonicity properties.
With binary-Leontief utilities, EDR is also egalitarian for projects and for agents (subject to decomposability), can be found efficiently using linear programming, and attained at the limit of a best-response sequence.
Buterin, Hitzig and Weyl[4] present a mechanism in which donors invest money to create public goods. They assume that agents have quasilinear utilities, so without coordination, there will be under-provision of public goods due to the free-rider problem.
They suggest a mechanism called Quadratic Finance, inspired by quadratic voting. The amount received by each project x is
\left(\sumi\sqrt{ci,x
Other ways to encourage public goods provision are:
They present variations and extensions of QF. They explain how it can be used to campaign finance reform, funding open source software, news media finance, charitable giving, and urban public projects.
Donor coordination with indivisible targets is similar to combinatorial participatory budgeting, except that in the latter, the budget is fixed in advance and not contributed voluntarily by the agents.
Aziz and Ganguly[5] study a variant on indivisible participatory budgeting in which there is no exogeneous budget. There is a list of potential projects, each with its own cost. Each agent approves a subset of the projects, and provides an upper bound on the amount of money he can donate. The utility of each agent equals the amount of money spent on projects he approves (i.e., cost-satisfaction). The rule should specify (1) Which projects are funded? (2) How much money each donor pays? Note that, because the projects are indivisible, probably most donors will pay less than their upper bound.
They study three axioms related to encouraging participation:
Three axioms related to efficiency:
Two axioms related to fairness:
Finally, they study strategyproofness. They study which axioms are satisfied by three welfare-maximization rules: utilitarian, egalitarian (leximin) and Nash-product; they also study their computational complexity. They also conduct experiments for studying the price of fairness - how much fairness properties effect the social welfare - in instances that model two real-life donor coordination scenarios: share-house setting, and crowdfunding setting.
Aziz, Gujar, Padala, Suzuki and Vollen[6] [7] extend the above study to agents with cardinal ballots and quasilinear utilities. They show that welfare maximization admits an FPTAS, but welfare maximization subject to a natural and weak participation requirement is strongly inapproximable.
Chen, Lackner and Maly[8] study an extension of indivisible participatory budgeting in which there is both exogeneous budget and potential donations. Each voter can, besides voting for projects, also donate to specific projects of his choice. The donations of each project are deducted from its cost before the PB rule is activated. Their aim is to guarantee that rich donors do not use their donations to have an unfairly large influence on the total budget. Formally, they define a condition called "Donation-No-Harm", which requires that the utility of each agent when there are donations is at least as high as his utility without donations. They also study monotonicity properties specific to the setting with donations. They assume cardinal utilities. They also assume that projects belong to possibly-overlapping categories, with upper and lower quotas on each category.
They study 8 rules: 4 based on global optimization and 4 based on greedy optimization. They consider three ways to adapt these rules to the setting with donations:
Besides Donation No Harm, they also study three monotonicity axioms: Donation-project-monotonicity, Donation-welfare-monotonicity, and Donation-voter-monotonicity. They also study two computational problems related to this setting:
In the Paris Declaration of 2005, donor countries agreed to coordinate their donations in order to eliminate duplication of efforts and better align foreign aid flows with priorities of the recipient countries. They acknowledged that aid fragmentation impairs the effectiveness of aid. However, Nunnenkamp, Ohler and Thiele[9] show that these ideas were not implemented in practice, and the donor coordination even declined. Leiderer[10] presents specific evidence for this from aid to the health and education sectors in Zambia.