Why Are You, an Effective Altruist, Donating to Charity?
How much should we save, and how much should we give?
One of the core beliefs that people in the Effective Altruist movement hold is a sense of egalitarianism across space – just because someone happened to be born in a different country does not mean that they are morally less important. Since we can save lives in Africa more cost effectively than we can in the United States, we donate across national borders.
It seems reasonable to think that this egalitarianism should extend across time as well. Suppose that in order to send a malaria net to someone in Africa, you needed to wait a month for it to arrive. It doesn’t seem reasonable to say that that person is worth less simply because they are a month into the future. So, I believe that we should regard those who are as being of the same intrinsic moral worth as those who will be.
Taking this seriously raises some troubling questions, though. If you discount future consumption at a rate lower than the market interest rate, why don’t you put the money which you would have donated into safe debt, collect the return, and donate it later? You maximize the good that you are doing in the world by doing something which is literally risk free. Donating money now smacks of selfishness. It might be the sort of noisy, self-serving goodness, like donating to an opera house, which effective altruism is a reaction against. Treating the present the same as the future requires patience and self-abnegation.
Of course, just because you are in principle indifferent between the present and future doesn’t mean that you won’t discount future consumption. The Ramsey equation for interest rate r consists of two terms: r equals the pure rate of time preference delta plus the coefficient of relative risk aversion eta times future growth g. The second term follows from people having curved utility functions, and so valuing gains relatively less when we’re richer. Being indifferent to present versus future consumption implies putting the delta to 0, but you still retain the second term. The argument for donating now, then, is that there won’t be poor people in the future.
But this, importantly, assumes homogenous agents. Very plausible modeling assumptions completely obviate this, and return the correct social discount rate to 0. Suppose that there is a distribution of people by income, with a floor representing subsistence. Income follows a Pareto distribution, with a long tail of increasingly poorer people. As incomes increase, it leads both to higher incomes, and to an increase in population because people who would have been censored by the subsistence floor are now able to live. I have not worked it out, but I am fairly sure that if there is a “constant elasticity of substitution” – i.e. the second derivative is a number – economic growth leads to no change in the absolute number of poor people.
If the poor are something we shall always have with us, then we can set up a model as follows. In each period, n people are born, who start at a consumption level of 0. People are infinitely lived, and exogenously increase their consumption by a constant in each period. Since the marginal utility of donating is constant, and we are indifferent between present and future consumption, we will want to donate the same amount in each period. The only way to ensure this is constant is to donate the interest without touching the principle.
The argument for non-zero discount rates may be familiar to us from debates over climate change. This differs from climate change debates because our aid can be targeted. We don’t need to care about the average person, just the very poorest. It is possible that economic growth will be broadbased enough that it isn’t possible to be as poor as the poorest in the world are now. Something like this has happened in the United States, where even being homeless is substantially better in terms of consumption than being poor in the Congo. You may be begging, but you’re begging in a place with AC everywhere, plentiful food, no endemic diseases, free clean water, and lots of other stuff. To the extent that we expect for economic growth to raise all members of society, we should want to donate more now.
The argument for saving instead of donating is robust, however, to assuming homogenous agents, so long as the market interest rate contains any amount of pure time preference. You would still hold a greater preference for delaying consumption than the average person, so you can profit by investing. Now, this pure time preference is extremely hard to sort out from pragmatically avoiding risks, as I argue in “Should We Take Everything from the Old to Give to the Young?”. People are uncertain about the system as a whole, or about their chances of dying, or like risks. If that is all there is to the pure rate of time preference delta, then no arbitrage is possible. If you, however, set up a trust so that the funds will not be redirected after you die, and people have less than perfect dynastic utility, then profitable arbitrage is still possible.
I want to preempt some possible arguments. Suppose that donations have economic spillovers beyond that of a simple transfer. This is something which recent studies have documented – see, for example, “General Equilibrium Effects of Cash Transfers”, which found that every dollar sent to Kenyan villages increased consumption by $2.58. This is, however, irrelevant to how we much we should donate. Only a change in growth rates would be relevant, because we would expect those spillovers to affect levels in the future too. If donations earlier create changes in growth rates, then it becomes akin to investing.
Another possible counterargument is that, since we will become richer later, there will be many more people willing to donate. If we say that “altruism” is a randomly drawn trait, and that this distribution doesn’t vary over time, then there is no reason to save more for the future. This hinges upon the Pareto distribution remaining stable over time, however. In the real world, the rich have lower fertility than the poor, and so there will be fewer rich people able to donate in the future. This militates for giving away one’s wealth later, rather than earlier.
If you have a belief that the world is going to radically change in the future, such that you cannot donate indefinitely, it is best to increase the proportion of donations to the portion saved so that the fund eventually expires. If you knew the exact number of periods which the world will go on for, you will set it so that it runs out with the last period; if you are uncertain, but have an expected value for the number of periods, then you will choose the same amount.
To summarize, whether you should give your money away now, or save it until later, depends on whether you believe that there will be poor people with the same marginal gain in consumption as those who exist now. Donations increase if you believe the world will radically change, and decrease if you do not. If you are unable to exhaust the number of poor people in the world, then it suggests waiting and saving. It also suggests that we should consider funding scientific inquiry more, relative to consumption transfers. Effective altruists should be more willing to hold on to savings so as to fund speculative ventures which they may have.
If you’re not already familiar with it, would be helpful to check out Phil Trammel’s work on Patient Philanthropy, which makes this/ a similar point.
One thing to add is that charitable opportunities can be unforseen and temporary, so saving money to spend on e.g. natural disasters is good.
In the end the solution might be something like having uncertainty across all parameters and allocating money to each time period proportional to the expected value (and updating as time goes on).