Eliezer, I think inflation caused via cash transfers results (under some fairly basic assumptions) in unchanged - not negative - total real wealth for the aggregate set of people experiencing the inflation, because this aggregate set of people includes the same set of people that causes the inflation as a result of having more currency. There may be situations in which "N people receive X units of currency, but the supply of goods they purchase remains fixed, so they experience inflation and do not end up with more real wealth", but not situations in which "N people receive X units of currency and as a result have less real wealth, or cause inflation for others that lowers the others' real wealth more than their own has risen."
If you believe that GiveDirectly's transfers cause negligible inflation for the people receiving them (as implied by the studies we've reviewed), this implies that those people become materially wealthier by (almost) the amount of the transfer. There may be other people along the chain who experience inflation, but these people at worst have unchanged real wealth in aggregate (according to the previous paragraph). (BTW, I've focused in on the implications of your scenario for inflation because we have data regarding inflation.)
It's theoretically possible that the distributive effects within this group are regressive (e.g., perhaps the people who sell to the GiveDirectly recipients then purchase goods that are needed by other lower-income people in another location, raising the price of those goods), but in order to believe this one has to make a seemingly large number of unjustified assumptions (including the assumption of an inelastic supply of the goods demanded by low-income Kenyans, which seems particularly unrealistic), and those distributive effects could just as easily be progressive.
It also seems worth noting that your concern would seem to apply equally to any case in which a donor purchases foreign currency and uses it to fund local services (e.g., from nonprofits), which would seem to include ~all cases of direct aid overseas.
If I'm still not fully addressing your point, it might be worth your trying a "toy economy" construction to elucidate your concern - something along the lines of "Imagine that there are a total of 10 people in Kenya; that 8 are poor, have wealth equal to X, and consume goods A and B at prices Pa and Pb; that 2 are wealthy, have wealthy equal to Y, and consume goods C and D at prices Pc and Pd. When the transfer is made, dollars are traded for T shillings, which are then distributed as follows, which has the following impact on prices and real consumption ..." I often find these constructions useful in these sorts of discussions to elucidate exactly the potential scenario one has in mind.
I first wrote up the following post, then happened to run into Holden Karnofsky in person and asked him a much-shortened form of the question verbally. My attempt to recount Holden's verbal reply is also given further below. I was moderately impressed by Holden's response because I had not thought of it when listing out possible replies, but I don't understand yet why Holden's response should be true. Since GiveWell has recently posted about objections to GiveDirectly and replies, I decided to go ahead and post this now.
A question for GiveWell:
Your current #2 top-rated charity is GiveDirectly, which gives one-time gifts of $1000 over 9 months, directly to poor recipients in Kenya via M-PESA.
Givewell tries for high standards of evidence of efficacy and cost-effectiveness. As I understand it, you don't just want the charity to be arguably cost effective, you want a very high probability that the charity is cost-effective.
The main evidence I've seen cited for direct giving is that the recipients who received the $1000 are then substantially better off 9 months later compared to people who aren't.
While I can imagine arguments that could repair the obvious objection to this reasoning, I haven't seen yet how the resulting evidence about cost-effectiveness could rise again to the epistemic standards one would expect of Givewell's #2 evidence-based charity.
The obvious objection is as follows: Suppose the Kenyan government simply printed new shillings and handed out $1000 of such shillings to the same recipients targeted by GiveDirectly. Although the recipients would be better off than non-recipients, this might not reflect any improvement in net utility in Kenya because no new resources were created by printing the money.
There are of course obvious replies to this obvious objection:
(1) Because the shillings handed out by GiveDirectly are purchased on the foreign currency exchange market using U. S. dollars, and would otherwise have been spent in Kenya in other ways, we should not expect any inflation of the shilling, and should expect an increase in Kenyan consumption of foreign goods corresponding to the increased price of shillings implied by GiveDirectly adding their marginal demand to the auction and thereby raising the marginal price of all shillings sold. The primary mechanism of action by which GiveDirectly benefits Kenya is by raising the price of shillings in the foreign exchange market and making more hard currency available to sellers of shillings. So far as I can tell, this argument ought to generalize: Any argument that the Kenyan government could not accomplish most of the same good by printing shillings will mean that the primary mechanism of GiveWell's effectiveness must be the U.S. dollars being exchanged for the shillings on the foreign currency market. This in turn means that GiveDirectly could accomplish most of its good by buying the same shillings on the foreign currency market and burning them.
(Or to sharpen the total point of this article: The sum of the good accomplished by GiveDirectly should equal:
Indeed, since these mechanisms of action seem mostly independent, we ought to be able to state a percentage of good accomplished which is allegedly attributed to each, summing to 1. E.g. maybe 80% of the good would be achieved by printing shillings and distributing them to the same recipients, and 20% would be achieved by purchasing shillings on the foreign exchange market and burning them. But then we have mostly the same questions as before about how to generate wealth by printing shillings.)
(2) Inequality in Kenya is such that redistributing the supply of shillings toward the very poor increases utility in Kenya. Thus the Kenyan government could accomplish as much good as GiveDirectly by printing an equivalent number of shillings and giving them to the same recipients. This would create inflation that is a loss to other Kenyans, some of them also very poor, but so much of the shilling supply is held by the rich that the net results are favorable. Printing shillings can create happiness because it shifts resources from making speedboats for the rich to making corrugated iron roofs for the poor.
(It would be nice if the Kenyan government just printed shillings for GiveDirectly to use, but this the Kenyan government will not realistically do. Effective altruists must live in the real world, and in the real world GiveDirectly will only accomplish its goals with the aid of effective altruists. One cannot live in the should-universe where Kenya's government is taking up the burden. Effective altruists should reason as if the Kenya government consists of plastic dolls who cannot be the locus of responsibility instead of them - that's heroic epistemology 101. Maybe there will eventually be returns on lobbying for Minimum Guaranteed Income in Kenya if the programs work, but that's for tomorrow, not right now.)
(3) Like the European Union, Kenya is not printing enough shillings under standard economic theory. (I have no idea if this is plausibly true for Kenya in particular.) If the government printed shillings and gave them to the same recipients, this would create real wealth in Kenya because the economy was operating below capacity and velocity of trade would pick up. The shillings purchased by GiveDirectly would otherwise have stayed in bank accounts rather than going to other Kenyans. Note that this contradicts the argument step in (1) where we said that the purchased shillings would otherwise have been spent elsewhere, so you should have questioned one argument step or the other.
(4) Village moneylenders and bosses can successfully extract most surplus generated within their villages by raising rents or demanding bribes. The only way that individuals can escape the grasp of moneylenders and rentiers is with a one-time gift that was not expected and which the moneylenders and bosses could not arrange to capture. The government could accomplish as much good as GiveDirectly by printing the same number of shillings and giving them to the same people in an unpredictable pattern. This would create some inflation but village moneylenders or bosses would ease off on people from whom they couldn't extract as much value, whereas the one-time gift recipients can purchase capital goods that will make them permanently better off in ways that don't allow the new value to be extracted by moneylenders or bosses.
If I recall correctly, GiveDirectly uses the example of a family using some of the gift money to purchase a corrugated iron roof. From my perspective the obvious objection is that they could just be purchasing a corrugated iron roof that would've gone to someone else and raising the prices of roofs. (1) says that Kenya has more foreign exchange on hands and can import, not one more corrugated iron roof, but a variety of other foreign goods; (2) says that the resources used in the corrugated iron roof would otherwise have been used to make a speedboat; (3) says that a new trade takes place in which somebody makes a corrugated iron roof that wouldn't have been manufactured otherwise; and (4) says that the village moneylenders usually adjust their interest rates so as to prevent anyone from saving up enough money to buy a corrugated iron roof.
The trouble is that all of these mechanisms of action seem much harder to measure and be sure of, than the measurable outcomes for gift recipients vs. non-recipients.
To reiterate, the sum of the good accomplished by GiveDirectly should equal the good accomplished by the Kenyan government printing shillings and distributing them to the same recipients, plus the good accomplished by GiveDirectly purchasing shillings on the foreign exchange market using US dollars and then burning them. It seems to me to be difficult to arrive at a state of strong evidence about either of the two terms in this sum, with respect to any mechanism of action I've thought of so far.
With respect to the second term in this sum: GiveDirectly buying shillings on the foreign exchange market and burning them might create wealth, but it's hard to see how you would measure this over the relevant amounts, and no such evidence was cited in the recommendation of GiveDirectly as the #2 charity.
With respect to the first term in this sum: Under the Bayesian definition of evidence, strong evidence is evidence we are unlikely to see when the theory is false. Even in the absence of any mechanism whereby printing nominal shillings creates happiness or wealth, we would still expect to find that the wealth and happiness of gift recipients exceeded the wealth of non-recipients. So measuring that the gift recipients are wealthier and happier is not strong or even medium evidence that printing nominal shillings creates wealth, unless I'm missing something here. Our posterior that printing shillings and giving them to certain people would create net wealth in any given quantity, should roughly equal our prior, after updating on the stated experimental evidence.
When I posed a shortened form of this question to Holden Karnofsky, he replied (roughly, I am trying to rephrase from memory):
Please keep in mind that this is Holden's off-the-cuff, non-written in-person response as rephrased by Eliezer Yudkowsky from imperfect memory.
With that said, I've thought about (what I think was) Holden's answer and I feel like I'm still missing something. I agree that if U.S. dollars were being sent directly to Kenyan recipients and used only to purchase foreign goods, so that foreign goods were being directly sent from the U.S. to Kenyan recipients, then improvement in measured outcome for recipients compared to non-recipients would be an appropriate metric, and that the decomposition would be perverse. But if the received money, in the form of Kenyan shillings, is being used primarily to purchase Kenyan goods, and causing those goods to be shipped to one villager rather than another while also possibly increasing velocity of trade, remedying inequality, and enabling completely different actors to buy some amount of foreign goods, then I honestly don't understand why this scenario should have the same causal mechanisms as the scenario where foreign goods are being shipped in from outside the country. And then I honestly don't understand why measured improvements for one Kenyan over another should be a good proxy for aggregate welfare change to the country.
I may be missing something that an economist would find obvious or I may have misunderstood Holden's reply. But to me, my sum seems like an obvious causal decomposition of the effects in Kenya, neither of whose terms can be estimated well. I don't understand why I should expect the uncertainty in these two estimates to cancel out when they are added; I don't understand what background causal model yields this conclusion.
To be clear, I personally would guess that the U.S. would be net better off, if the Federal Reserve directly sent everyone in the U.S. with income under $20K/year a one-time $6,000 check with the money phasing out at a 10% rate up to $80K/year. This is because, in order of importance:
What I wouldn't say is that my belief in the above is as strong as my belief in, say, the intelligence explosion. I'd guess that the printing operation would do more good than harm, but it's not what I would call a strong evidence-based conclusion. If we're going to be okay with that standard of argument generally, then the top charity under that standard of reasoning, generally and evenhandedly applied, ought to work out to some charity that does science and technology research. (X-risk minimization might seem substantially 'weirder' than that, but the best science-funding charities should be only equally weird.) And I wouldn't measure the excess of happiness of gift-recipients compared to non-recipients in a pilot program, and call this a good estimate of the net good if a Minimum Guaranteed Income were universally adopted.
So to reiterate, my question to Givewell is not "Why do you think GiveDirectly might maybe end up doing some good anyway?" but "Does GiveDirectly rise to the standards required for your #2 evidence-based charity?"