“Why Nations Fail” and the long-termist view of global poverty

Link post

Within the effec­tive al­tru­ism com­mu­nity, peo­ple of­ten talk about “long-ter­mist” vs “short-ter­mist” wor­ld­views. The offi­cial dis­tinc­tion be­tween the two is that short-ter­mists pri­ori­tize prob­lems by how they af­fect peo­ple al­ive to­day, while long-ter­mists pri­ori­tize prob­lems by how they could af­fect hu­man­ity’s en­tire fu­ture tra­jec­tory. In prac­tice, peo­ple usu­ally treat this as syn­ony­mous with pri­ori­tiz­ing ei­ther ex­is­ten­tial risk re­duc­tion (if long-ter­mist), or scal­ing up proven global health in­ter­ven­tions (if short-ter­mist).

It’s a bit sur­pris­ing that each wor­ld­view should have ex­actly one fa­vorite cause area, though. Couldn’t you have short-ter­mist work on ex­is­ten­tial risk, or long-ter­mist work on global poverty? In re­al­ity, these sup­pos­edly dis­crete wor­ld­views seem more like cor­re­lated clusters of var­i­ous differ­ent be­liefs:


  • High time dis­count rate

  • Prefers highly ro­bust “out­side view” type arguments

  • Ex­trap­o­lates ex­ist­ing effects or trends

  • Skep­ti­cal of prima fa­cie bizarre claims

  • Fo­cuses on fix­ing known, con­crete problems

  • Fast feed­back loops are crit­i­cal to mak­ing progress


  • Low or no dis­count rate

  • More open to “in­side view” that this case might be different

  • Rea­sons about the fu­ture from first principles

  • Takes weird-sound­ing ideas more seriously

  • Fo­cuses on pre­vent­ing hy­po­thet­i­cal, neb­u­lous risks

  • Fast feed­back is helpful, but not the most im­por­tant thing

It’s un­der­stand­able why some of these are cor­re­lated, but there must be a lot of peo­ple who fall through the cracks be­tween the clusters. What if you share short-ter­mists’ skep­ti­cism of weird claims and hy­po­thet­i­cal risks, but you’re will­ing to fo­cus on first-prin­ci­ples rea­son­ing and work on a long time scale?

You’d still want to fo­cus on some­thing that’s a prob­lem to­day, so you’d prob­a­bly want to work on global poverty. But you’d dis­miss GiveWell’s top char­i­ties as treat­ing a symp­tom and not a cause. Why do these coun­tries need char­ity in the first place? South Korea used to be just as poor as any­where in Africa, but to­day it’s in­cred­ibly pros­per­ous, while sub-Sa­haran Africa has made way less progress. If we could move the low­est-growth coun­tries from their cur­rent tra­jec­tory onto South Korea’s, we’d have done much more than any sin­gle malaria-erad­i­ca­tion cam­paign could.

If that were your wor­ld­view, you’d re­ally en­joy Why Na­tions Fail, one of the best at­tempts I’ve seen at get­ting a first-prin­ci­ples un­der­stand­ing of what af­fects coun­tries’ long-term eco­nomic growth.

First of all, what is eco­nomic growth? It’s when peo­ple pro­duce more (or more valuable) stuff with the same effort[1]. The first and least con­tro­ver­sial point in Why Na­tions Fail is that for a na­tion to keep on do­ing more with less, its in­di­vi­d­ual cit­i­zens need to be in­cen­tivized to be­come more pro­duc­tive. In par­tic­u­lar, the state should not set up sys­tems where, when­ever some­one gets more pro­duc­tive, other peo­ple come and take away the ex­tra stuff they pro­duced. Those sys­tems are what the au­thors call ex­trac­tive eco­nomic in­sti­tu­tions, and they in­clude things like slav­ery, serf­dom, in­den­tured servi­tude, rov­ing ban­dits, guilds, col­lec­tivized agri­cul­ture, na­tion­al­iza­tion of pri­vate as­sets, offi­cials re­quiring bribes, kan­ga­roo courts, ba­nana re­pub­lics, and other [an­i­mal or veg­etable] [civic in­sti­tu­tion].

You might think you could grow your econ­omy un­der ex­trac­tive in­sti­tu­tions by forc­ing peo­ple to be­come more pro­duc­tive even if they won’t get to keep the sur­plus. This does of­ten work in the short term (and the short term can last a sur­pris­ingly long time)–for in­stance, Soviet Rus­sia grew at about 5% an­nu­ally from 1930-1970,[2] de­spite hav­ing ex­tremely ex­trac­tive in­sti­tu­tions. But that only worked be­cause the growth started from a low base. Soviet Rus­sia’s eco­nomic in­sti­tu­tions, while hor­rible, were ar­guably still less ex­trac­tive than the serf­dom that pre­ceded them. Plus, the growth came mainly from “easy wins” like peo­ple switch­ing from hor­rifi­cally in­effi­cient farm­ing to hor­rifi­cally in­effi­cient in­dus­try.

In fact, growth un­der ex­trac­tive in­sti­tu­tions can only come from easy wins–easy enough that they’re ob­vi­ous to who­ever has the job of forc­ing other peo­ple to be more pro­duc­tive. (No­body is go­ing to gen­er­ate their own pro­duc­tivity-boost­ing ideas, be­cause they have no in­cen­tive to.) Any half-de­cent Soviet cen­tral plan­ner could have told you that in­dus­try would be more pro­duc­tive than farm­ing, so they were able to force peas­ants to make the switch. But no­body was good enough at cen­tral plan­ning to fix all the ways their in­dus­try was in­effi­cient–let alone come up with new in­ven­tions or tech­nolo­gies as fast as the US was. So once all the low-hang­ing fruit was plucked, circa 1980, the Soviet Union ran into a pro­duc­tivity wall and col­lapsed.

What’s more, even if an in­no­va­tion is ob­vi­ous to the ex­trac­tive elite, they might ac­tively try to crush it be­cause it threat­ens their hold on power. For in­stance, in 19th-cen­tury Europe, it was ob­vi­ous that fac­to­ries and railroads could be a huge pro­duc­tivity im­prove­ment, but the ex­trac­tive rulers of periph­eral Europe wanted none of them. Here’s Aus­tria:

When a plan to build a north­ern railway was put be­fore [Em­peror] Fran­cis I, he replied, “No, no, I will have noth­ing to do with it, lest the rev­olu­tion might come into the coun­try.”

And here’s Rus­sian Fi­nance Minister (1823-44) Count Igor Kankrin:

[R]ailways do not always re­sult from nat­u­ral ne­ces­sity, but are more an ob­ject of ar­tifi­cial need or lux­ury. They en­courage un­nec­es­sary travel from place to place, which is en­tirely typ­i­cal of our time.

It might seem car­toon­ish for a ruler to ban tech­nol­ogy out of fear of TEH REVOLUTIONZ–un­til you re­mem­ber how many ex­trac­tive states to­day, like Ethiopia and China, have re­sisted the In­ter­net for similar rea­sons.

The next part of Why Na­tions Fail asks: if eco­nomic in­sti­tu­tions ex­plain long-term growth, what ex­plains long-term eco­nomic in­sti­tu­tions? The an­swer, maybe ob­vi­ously, is poli­tics.

Ex­trac­tive eco­nomic in­sti­tu­tions tend to per­sist more when so­cial power is con­cen­trated in a small elite (“ex­trac­tive poli­ti­cal in­sti­tu­tions”) that can use this power to take the eco­nomic sur­plus for them­selves–or just de­stroy it if it looks scary and po­ten­tially desta­bi­liz­ing. They per­sist less when so­cial power is dis­tributed widely in the pop­u­la­tion (“in­clu­sive poli­ti­cal in­sti­tu­tions”) be­cause in that case the peo­ple be­ing ex­tracted will of­ten have enough poli­ti­cal power to fight back against the ex­trac­tion.

For in­stance, in the early 1900s in the United States, var­i­ous in­dus­tries (steel, oil, bank­ing, chem­i­cals, farm­ing tools, etc.) ended up un­der the con­trol of ex­trac­tive mo­nop­o­lists who charged too much, paid too lit­tle, pock­eted the differ­ence, and spent the differ­ence on gold-plated toi­lets. But even the world’s fan­ciest toi­lets couldn’t defeat the mass pop­ulist move­ment that rose up to op­pose the mo­nop­o­lies (be­cause the US was suffi­ciently plu­ral­is­tic that mass pop­ulist move­ments were very pow­er­ful), and so they were even­tu­ally forcibly bro­ken up and reg­u­lated.

In states with less plu­ral­is­tic poli­tics, the elites can get away with this kind of ex­trac­tion. For in­stance, when the Bri­tish colonized Sierra Leone, they set up monopoly co­coa and coffee pur­chas­ing boards that bought low from the farm­ers, sold high on in­ter­na­tional mar­kets, and pock­eted the mar­gin. After in­de­pen­dence, in­stead of dis­man­tling the boards, Sierra Leone’s pres­i­dents turned them into a slush fund, driv­ing the mar­gin up to 90%.

Sierra Leone illus­trates an­other sad fact about ex­trac­tive eco­nomic in­sti­tu­tions, which is that they per­sist sur­pris­ingly much across nom­i­nally differ­ent poli­ti­cal regimes. What hap­pened in Sierra Leone hap­pened nearly ev­ery­where in sub-Sa­haran Africa: the colon­ists were ousted by a rev­olu­tion­ary leader, who quickly re­al­ized that the colon­ists’ gi­ant money-print­ing ma­chine was still in place, and de­cided that maybe the en­tire colo­nial ap­para­tus didn’t need to be smashed. And of course, in or­der to keep their hold on the eco­nomic sur­plus, the new “demo­cratic” gov­ern­ment quickly re­sorted to re­pres­sion and vi­o­lence. (Why Na­tions Fail is very clear that what mat­ters is de facto, not de jure, democ­racy, and that con­fus­ing these two will lead to very wrong con­clu­sions.)

Over long timescales, then, it seems like a mix of in­clu­sive and ex­trac­tive in­sti­tu­tions is un­sta­ble. If in­sti­tu­tions are in­clu­sive enough, then any group that de­vi­ates and tries to in­sti­tute ex­trac­tive norms will get shut down by the rest of so­ciety. If in­sti­tu­tions aren’t in­clu­sive enough, one of those groups will even­tu­ally suc­ceed, and then be in­cen­tivized to con­soli­date their power even fur­ther un­til the so­ciety is to­tally ex­trac­tive.

The most glar­ing ex­cep­tion to this rule–which many re­view­ers have pointed out–is China since Deng Xiaop­ing, which has rel­a­tively in­clu­sive eco­nomic in­sti­tu­tions, but ex­tremely au­thor­i­tar­ian poli­tics, for the last 40 years. Ace­moglu and Robin­son ac­knowl­edge this, but claim, ba­si­cally, that China is not in a sta­ble equil­ibrium:

All the same, [Chi­nese] growth will run out of steam un­less ex­trac­tive poli­ti­cal in­sti­tu­tions make way for in­clu­sive in­sti­tu­tions. As long as poli­ti­cal in­sti­tu­tions re­main ex­trac­tive, growth will be in­her­ently limited, as it has been in all other similar cases.

This is a kind of weak ex­pla­na­tion as writ­ten be­cause they don’t even men­tion the mechanism by which it will be limited, but with some char­ity I think you can con­struct a de­cent ar­gu­ment. Con­cretely, it seems like the _Why Na­tions Fail_model pre­dicts that China’s cur­rent, in­clu­sive eco­nomic in­sti­tu­tions are not sus­tain­able with­out in­clu­sive poli­ti­cal in­sti­tu­tions to sup­port them. In the cur­rent regime, the Chi­nese poli­ti­cal elite’s lo­cal in­cen­tives are to ex­tract as much of the econ­omy’s sur­plus as pos­si­ble, be­cause they can; it’s not sus­tain­able to just hope they keep be­ing al­tru­is­tic enough to not do that. At some point, un­less China’s poli­ti­cal in­sti­tu­tions give real power to a broad ma­jor­ity of peo­ple, the elites will give in to temp­ta­tion and start tak­ing all the eco­nomic sur­plus for them­selves.

(Note by the way that China’s re­cent eco­nomic slow­down does not con­firm this model’s pre­dic­tions, un­less you think the slow­down was caused by the poli­ti­cal elite be­com­ing more eco­nom­i­cally ex­trac­tive, which doesn’t seem like the case.)

If you’re wor­ried about hav­ing pow­er­ful au­thor­i­tar­ian na­tions hang­ing around, this is a com­fort­ing model to have. On the other hand, China has been “out of equil­ibrium” for al­most half a cen­tury, and Why Na­tions Fail doesn’t offer any frame­work for think­ing about how long these dis­e­quil­ibria can last.

The strongest claim in Why Na­tions Fail, which is mostly left an­noy­ingly im­plicit, is that in­sti­tu­tions are the dom­i­nant fac­tor in shap­ing eco­nomic growth, over long enough time hori­zons. Many of their re­view­ers dis­puted this.

To be fair, a lot of those re­view­ers seemed to have rounded off this claim to “in­sti­tu­tions are always the dom­i­nant fac­tor,” and then com­plained about the book not ex­plain­ing China (de­spite the half-chap­ter ex­plain­ing that China is out of equil­ibrium), or not suffi­ciently ex­plain­ing Bo­livia vs Viet­nam over the last 30 years (if China can be out of equil­ibrium for 40 years, so can other coun­tries), or similar. But even over long pe­ri­ods of time, it’s rea­son­able to ask whether in­sti­tu­tions were re­ally the most im­por­tant pos­si­ble thing.

The most ob­vi­ous al­ter­na­tive can­di­date is ge­og­ra­phy (in­clud­ing nat­u­ral re­sources). Why Na­tions Fail makes some at­tempt to ar­gue that ge­og­ra­phy isn’t suffi­ciently ex­plana­tory, cit­ing (a) the di­ver­gence be­tween North and South Korea, and (b) that af­ter Euro­pean coloniza­tion of the Amer­i­cas, the cor­re­la­tion be­tween lat­i­tude and pro­duc­tivity went from nega­tive (trop­ics more pro­duc­tive) to pos­i­tive (tem­per­ate re­gions more pro­duc­tive):

As we saw in the last chap­ter, at the time of the con­quest of the Amer­i­cas by Colum­bus, the [trop­i­cal] ar­eas… held the great Aztec and Inca civ­i­liza­tions.… In sharp con­trast… the [mod­ern] United States, Canada, Ar­gentina, and Chile, were mostly in­hab­ited by Stone Age civ­i­liza­tions lack­ing these tech­nolo­gies. The trop­ics in the Amer­i­cas were thus much richer than the tem­per­ate zones, sug­gest­ing that the “ob­vi­ous fact” of trop­i­cal poverty is nei­ther ob­vi­ous nor a fact. In­stead, the greater riches in the United States and Canada rep­re­sent a stark re­ver­sal of for­tune rel­a­tive to what was there when the Euro­peans ar­rived.

The re­ver­sal of for­tune hap­pened, they ar­gue, be­cause colon­ists were more likely to in­stall ex­trac­tive in­sti­tu­tions in the densely pop­u­lated (wealthiest) ar­eas of the Amer­i­cas, where they could rely on forced in­dige­nous la­bor; and in­clu­sive in­sti­tu­tions in the sparsely pop­u­lated ar­eas where there was no forced la­bor sup­ply and set­tlers needed to be in­cen­tivized to be pro­duc­tive. This differ­ence in in­sti­tu­tions even­tu­ally led to stag­na­tion in the trop­ics and faster growth in the tem­per­ate re­gions.

This doesn’t seem like a very defini­tive re­fu­ta­tion, though. The Koreas ex­am­ple only shows that in­sti­tu­tions dom­i­nate when ge­og­ra­phy and cul­ture are held con­stant (plus, it’s cherry-picked to be max­i­mally ex­treme). The ex­am­ple of the Amer­i­cas is stronger, but I can imag­ine al­ter­na­tive mod­els. For in­stance, maybe be­ing trop­i­cal was an ad­van­tage be­fore the in­dus­trial rev­olu­tion be­cause it meant less lethal weather; but a dis­ad­van­tage af­ter­wards be­cause it meant that your com­par­a­tive ad­van­tage was farm­ing, rather than in­dus­try, and farm­ing-spe­cial­ized economies grow less quickly be­cause they don’t de­velop com­pound­ing tech­ni­cal knowl­edge.

The ex­am­ple of the Amer­i­cas’ re­ver­sal of for­tune also doesn’t ad­dress more com­plex ge­o­graph­i­cal fac­tors, like that iron, coal or nav­i­gable rivers are nec­es­sary for in­dus­trial growth. In the medium term, ge­og­ra­phy of this form seems ob­vi­ously very im­por­tant; it’s no co­in­ci­dence that England was rich in all three. In the long term, it seems like they might be­come less im­por­tant, be­cause the types of re­sources that are im­por­tant will change. (To­day if you’re a de­vel­op­ing coun­try with no iron or coal de­posits, you can trade with other coun­tries to get them; if you don’t have nav­i­gable rivers, you can build railroads or high­ways.) But nei­ther of those is a perfect sub­sti­tute, so it seems like there’s still a strong case for ge­og­ra­phy mat­ter­ing.

On the other hand, if you’re try­ing to en­courage de­vel­op­ment, not just un­der­stand it, it’s not clear how ac­tion­able the pro-ge­og­ra­phy ar­gu­ment is. Even if ge­og­ra­phy is im­por­tant, it’s not par­tic­u­larly tractable to change! The only im­pli­ca­tion I can think of is maybe that liber­al­iz­ing im­mi­gra­tion be­comes more im­por­tant, so that it’s eas­ier for peo­ple to move from re­source-poor to re­source-rich ar­eas.

If true, the claims in Why Na­tions Fail have huge im­pli­ca­tions for (global-poverty-fo­cused) effec­tive al­tru­ism. Global poverty EA cur­rently fo­cuses on scal­ing up de­vel­op­ment aid in­ter­ven­tions that have a strong ev­i­dence base of ran­dom­ized con­trol­led tri­als. But if you buy Why Na­tions Fail‘s ar­gu­ment, you should prob­a­bly pre­fer to take coun­tries with ex­trac­tive in­sti­tu­tions and move them to­wards in­clu­sivity, if that prob­lem is tractable.

To me, it does seem tractable. For in­stance, in the re­cent elec­tion in Sene­gal, where I live, the in­cum­bent pres­i­dent had a huge spend­ing ad­van­tage, which was widely ru­mored to be be­cause he di­verted pub­lic funds to his cam­paign. (Not co­in­ci­den­tally, many promi­nent op­po­si­tion poli­ti­ci­ans are cur­rently in jail for mi­suse of pub­lic funds–the re­sult of an in­ves­ti­ga­tion ini­ti­ated by the same in­cum­bent pres­i­dent.) The cur­rent Sene­galese me­dia is re­luc­tant to in­ves­ti­gate be­cause the me­dia in the cap­i­tal are largely owned by peo­ple high up in gov­ern­ment. Spon­sor­ing the de­vel­op­ment of in­de­pen­dent in­ves­tiga­tive me­dia in Sene­gal seems like a (rel­a­tively) straight­for­ward step to­wards im­prov­ing Sene­galese poli­ti­cal in­sti­tu­tions.

Of course, this take is only based on a cou­ple hours of con­ver­sa­tion with my Sene­galese cowork­ers, so it’s definitely not op­ti­mal. Maybe it would even make things worse! But hope­fully it illus­trates the gen­eral point, which is that we know what in­clu­sive in­sti­tu­tions look like, so try­ing to move things in that di­rec­tion is mostly a ques­tion of strat­egy and ex­e­cu­tion to­wards a known goal, not of solv­ing some sort of mys­tery.

Why Na­tions Fail pro­vides a com­pel­ling the­ory, with clear im­pli­ca­tions, sup­ported by many ex­am­ples, a few of which (like the Koreas or Amer­i­cas ex­am­ples) even had some kind of “con­trol group.” But as the book goes on, it be­comes more and more glar­ing how much it re­lies on anec­dotes, not stud­ies. Un­for­tu­nately, that’s prob­a­bly be­cause the em­piri­cal situ­a­tion is abysmal.

Ace­moglu and Robin­son might not agree with that claim, since they were pub­lish­ing em­piri­cal stud­ies of in­sti­tu­tions in eco­nomics jour­nals for a long time be­fore they wrote Why Na­tions Fail. But they also chose not to cover most of that em­piri­cal re­search in the book, and I can see why: I couldn’t find a sin­gle study that seemed re­ally defen­si­ble.

Some of the prob­lems I saw were silly and avoid­able, like treat­ing Lik­ert scales as con­tin­u­ous in­stead of lev­eled-cat­e­gor­i­cal or ob­vi­ously vi­o­lat­ing the as­sump­tions of in­stru­men­tal vari­ables. But as far as I could tell, the big­ger is­sues were core to any at­tempt to em­piri­cally study the causes of growth:

  • The sam­ple size is tiny. There are only 200 coun­tries, many of which are too small or too weird to be in­for­ma­tive. (We do have hun­dreds of years of growth data for some coun­tries, but differ­ent years of growth for the same coun­try aren’t in­de­pen­dent data points.) The “effec­tive” sam­ple size is even lower be­cause the “er­ror term” is of­ten spa­tially cor­re­lated, which most stud­ies do not prop­erly ac­count for, caus­ing wide­spread p-value in­fla­tion.

  • When you’re an­a­lyz­ing coun­tries’ growth, ev­ery­thing is en­doge­nous. Your in­sti­tu­tions in year 1 af­fect your growth in year 2, which af­fects who gets more power in year 3, which af­fects how your in­sti­tu­tions change in year 4. If you mea­sure a cor­re­la­tion be­tween growth and in­sti­tu­tions, it could just as eas­ily be that fast growth causes bet­ter in­sti­tu­tions, not the other way around.

  • In­sti­tu­tions and growth have a re­ally com­pli­cated causal struc­ture–both of them have a huge num­ber of causes and effects. In situ­a­tions like that, and es­pe­cially with a sam­ple size of 200, cor­re­la­tional stud­ies never work. Even if you try to con­trol for all the con­founders, there are just too many sub­tle things that can go wrong.

  • Or­di­nar­ily you might try to get around that by us­ing a fancier causal in­fer­ence tech­nique, like find­ing in­stru­men­tal vari­ables–ran­dom events that af­fect a state’s in­sti­tu­tions, but aren’t cor­re­lated with them. For ex­am­ple, one fa­mous pa­per tried to do this with Euro­pean mor­tal­ity rates in colonized coun­tries (the the­ory be­ing that colon­ists were less likely to in­stall ex­trac­tive in­sti­tu­tions in coun­tries where more Euro­peans had im­mi­grated).
    But a key as­sump­tion of in­stru­men­tal vari­ables is that the in­stru­ment only af­fects the out­come (growth) via its effect on other things you’ve mea­sured (in­sti­tu­tions or other co­vari­ates). In other words, to get a valid es­ti­mate, you’d need to mea­sure ev­ery pos­si­ble way that set­tler mor­tal­ity could af­fect growth rates. Since mea­sur­ing ev­ery­thing is im­pos­si­ble, and throw­ing it all into a model with only 200 data points is also im­pos­si­ble, that makes in­stru­men­tal vari­ables dou­ble-im­pos­si­ble. (Which didn’t pre­vent var­i­ous economists from try­ing–ap­par­ently badly.)

Th­ese prob­lems make me put al­most zero weight on the em­piri­cal liter­a­ture for in­sti­tu­tions and growth, even be­fore read­ing about the spe­cific prob­lems with spe­cific pa­pers. But they ap­ply equally well to any at­tempt to find the causes of eco­nomic growth.

So where does that leave us?

Partly, it leaves us with a big method­olog­i­cal gap. The prob­lems above make it rel­a­tively harder to get to a true un­der­stand­ing of what mat­ters for de­vel­op­ment. But it doesn’t mean it’s hope­less to try and make in­tel­lec­tual progress on the ques­tion.

In­deed, I can imag­ine analy­ses that would make me more con­fi­dent that the in­sti­tu­tional hy­poth­e­sis is cor­rect–for in­stance, a sys­tem­atic re­view with case stud­ies of many differ­ent growth epi­sodes, or many differ­ent in­sti­tu­tional changes that you’d ex­pect to af­fect growth. I’m not sure why Ace­moglu and Robin­son’s em­piri­cal work fo­cused in­stead on what looks like badly-val­i­dated in­stru­men­tal-vari­ables mod­els–maybe be­cause they were eas­ier to do (or eas­ier to pub­lish in top jour­nals)? But given the amount of noise and the limited sam­ple size, case stud­ies seem much more likely to be in­for­ma­tive than fancy mod­els with ques­tion­able as­sump­tions.

Nor is it un­prece­dented for lots of ex­perts to agree on some­thing even with­out air­tight stud­ies back­ing it up. A real-world role model could be some­thing like the laws of sup­ply and de­mand–which most economists be­lieve in, even though they turn out to be tricky to ver­ify in many real-world cases, like min­i­mum wage laws or hous­ing. Or even like the effi­cient-mar­ket hy­poth­e­sis, which many effec­tive al­tru­ists seem to en­dorse, far be­yond our abil­ity to em­piri­cally test it.

But on ques­tions of poverty re­duc­tion, the effec­tive al­tru­ism com­mu­nity doesn’t seem to be re­ally look­ing for any method­ol­ogy other than ran­dom­ized con­trol­led tri­als. In fact, this is a ma­jor (and not of­ten ad­dressed) crit­i­cism of global-poverty-fo­cused effec­tive al­tru­ism, in­clud­ing from Ace­moglu him­self in his_Bos­ton Re­view_ cri­tique of effec­tive al­tru­ism:

[P]re­cise mea­sure­ment of the so­cial value of a donated dol­lar may be im­pos­si­ble. … If, as some economists and poli­ti­cal sci­en­tists sug­gest, changes in poli­ti­cal and eco­nomic in­sti­tu­tions are crit­i­cal for long-run eco­nomic growth, then watch­dog or­ga­ni­za­tions such as Amnesty may be es­sen­tial for trans­form­ing dys­func­tional regimes. Effec­tive al­tru­ists don’t (yet?) see the im­por­tance of these more poli­ti­cal or­ga­ni­za­tions. If this nar­row fo­cus con­tinues, it may di­vert pub­lic and me­dia pri­ori­ties from poli­ti­cal fac­tors un­der­pin­ning eco­nomic de­vel­op­ment.

His con­cern was shared by An­gus Deaton:

More broadly, the ev­i­dence for de­vel­op­ment effec­tive­ness, for “what works,” mostly comes from the re­cent wave of ran­dom­ized ex­per­i­ments. … How can those ex­per­i­ments be wrong? Be­cause they con­sider only the im­me­di­ate effects of the in­ter­ven­tions, not the con­texts in which they are set. Nor, most im­por­tantly, can they say any­thing about the wide-rang­ing un­in­tended con­se­quences.
How­ever coun­ter­in­tu­itive it may seem, … [d]evel­op­ment is nei­ther a fi­nan­cial nor a tech­ni­cal prob­lem but a poli­ti­cal prob­lem, and the aid in­dus­try of­ten makes the poli­tics worse.

Peter Singer’s re­sponse (which, as I’ve ar­gued be­fore, misses the point):

[I]f large-scale re­form offers some prospect of re­duc­ing poverty, then effec­tive al­tru­ists will try to as­sess its chance of do­ing good, and if the ex­pected value of such ac­tion is higher than the ex­pected value of more limited in­ter­ven­tions, they will ad­vo­cate work­ing for the large-scale re­forms.

It’s been four years since Peter Singer wrote that, and I don’t see much sign of the pre­dic­tion com­ing true. The lead­ing EA global poverty char­ity eval­u­a­tor, GiveWell, begged off be­cause in­sti­tu­tion-fo­cused in­ter­ven­tions wouldn’t be trans­par­ent enough:

Root-causes-based ap­proaches are, in our view, the kind of spec­u­la­tive and long-term un­der­tak­ings that are best suited to highly en­gaged donors (as dis­cussed above).

But the EA com­mu­nity doesn’t seem to have pro­duced many such “highly en­gaged donors,” with the pos­si­ble ex­cep­tion of Aceso Un­der Glass giv­ing to Tostan. (GiveWell it­self may be in the pro­cess of be­com­ing one. I hope that changes this!) In­stead, most global-poverty-fo­cused effec­tive al­tru­ists seem happy to fol­low GiveWell’s recom­mended char­i­ties. Maybe that’s be­cause all the long-ter­mists are work­ing on miti­gat­ing ex­is­ten­tial risks. But for any who are still sym­pa­thetic to global poverty as a cause area, Why Na­tions Fail pro­vides an in­ter­est­ing ex­am­ple of an al­ter­nate paradigm.

Thanks to Eve Bi­gaj, Drew Durbin, Milan Griffes, Alexey Guzey, Holden Karnofsky, Joyce Keeley, Jeff Kauf­man, Dan Luu, Aaron T, and Yuri Vish­nevsky for their com­ments on drafts of this post.

  1. Tech­ni­cally eco­nomic growth also in­cludes in­creas­ing the amount of effort–for in­stance, via pop­u­la­tion growth or via in­creas­ing hours worked–but those are much bet­ter-un­der­stood, so we’ll only think about pro­duc­tivity growth here. ↩︎

  2. Source: the most con­ser­va­tive es­ti­mates on this chart. ↩︎