James Snowden: The Evolution of GiveWell’s Research

Since 2007, GiveWell has recom­mended char­i­ties that di­rectly de­liver in­ter­ven­tions to benefi­cia­ries, with clear ad­van­tages that are rel­a­tively easy to mea­sure. Re­cently, it be­gan look­ing into in­ter­ven­tions that are harder to mea­sure — but might be much more cost-effec­tive.

Below is a tran­script of the talk, which we have lightly ed­ited for clar­ity. You may also watch it on YouTube or read it on effec­tivealtru­ism.org.

The Talk

My name is James. I’m a re­searcher at GiveWell and I’m go­ing to provide a quick up­date on some of the work we’re cur­rently do­ing to in­ves­ti­gate giv­ing op­por­tu­ni­ties re­lated to policy. [Since most of you are fa­mil­iar with GiveWell’s pre­vi­ous work, I won’t] go over what GiveWell has done in the past.

Slide02

First I’ll talk about how and why GiveWell is chang­ing. Then I’ll cover our cur­rent work, with a fo­cus on policy and, in par­tic­u­lar, pub­lic health reg­u­la­tion. Fi­nally, I’ll talk about what this means for GiveWell over the long term, and also what it might mean for [you].

GiveWell’s mis­sion is to find the best giv­ing op­por­tu­ni­ties we pos­si­bly can to help the global poor. We pub­lish our re­search and ra­tio­nale on­line, so donors can [more ju­di­ciously] choose where to di­rect their funds.

Slide03

In the past, we’ve gen­er­ally fo­cused on di­rect de­liv­ery [of in­ter­ven­tions], fund­ing or­ga­ni­za­tions that do work like bed­net dis­tri­bu­tion, cash trans­fers, or Vi­tamin A sup­ple­men­ta­tion. That is be­cause the two peo­ple who founded GiveWell in 2007 had very lit­tle ex­pe­rience in the global health and de­vel­op­ment space. There­fore, they wanted to work in ar­eas that were eas­ier to quan­tify and com­pare; this also meant that donors could trust GiveWell’s judg­ment [as it was based largely on calcu­la­tion rather than sub­jec­tive eval­u­a­tion].

While there are still a lot of judg­ment calls in­volved in de­cid­ing which di­rect-de­liv­ery giv­ing op­por­tu­ni­ties are the most cost-effec­tive, I think it is gen­er­ally a more ob­jec­tive pro­cess than se­lect­ing policy-re­lated op­por­tu­ni­ties. Tack­ling policy work re­quires ad­dress­ing some thorny ques­tions.

In the past year, two things have changed for GiveWell:

1. The or­ga­ni­za­tion has been do­ing this work for twelve years now, so we have a lot more ex­pe­rience in global health and de­vel­op­ment. We’ve re­cently hired peo­ple who are well-equipped to help us with [policy] work.

2. We have built up a level of trust with donors that al­lows us to make more sub­jec­tive judg­ment calls; they have rea­son to be­lieve that we’re not just talk­ing non­sense.

So what does this mean in terms of the ar­eas that are now within our scope? One way of think­ing about the spaces [we’re ex­plor­ing] is to carve it into three ar­eas:

Slide04



1. Policy work, which I’ll talk about to­day.

2. Direct in­ter­ven­tions we pre­vi­ously hadn’t con­sid­ered. Direct in­ter­ven­tions with less ex­per­i­men­tal ev­i­dence might, for ex­am­ple, in­clude those in the ed­u­ca­tion sec­tor — where it is a lot harder to draw con­clu­sions than it is in the health sec­tor be­cause [it’s hard to gen­er­al­ize re­sults from one lo­ca­tion to other lo­ca­tions]. Pre­vi­ously, that caused us to drop that re­search and de­ter­mine that we couldn’t come to strong con­clu­sions. Now we are ask­ing, “What is our best guess? How can we make rea­son­able judg­ments in ar­eas that are harder to eval­u­ate?”

3. Science. We haven’t made a lot of progress in this area yet, but we are con­sid­er­ing which tech­nolo­gies in the field of sci­ence we think could have a large im­pact on the global poor. For ex­am­ple, we are look­ing into whether there are op­por­tu­ni­ties to fund im­proved di­ag­nos­tic tools for pneu­mo­nia.

With policy as the fo­cus of this talk, I’ll give a brief overview of why [GiveWell de­cided to look into it].

Slide05

As a gen­eral caveat up front, I have only 20 min­utes to talk about an en­tire year’s work, so I’m not go­ing to get into as much de­tail as you would typ­i­cally ex­pect from GiveWell. We also haven’t yet writ­ten pub­li­cly about this work.

So why [fo­cus on] policy? Broadly, there are two rea­sons why we think this might be an area in which we can find bet­ter giv­ing op­por­tu­ni­ties:

1. There is a very strong in­tu­itive case for it. A small amount of fund­ing that is [ap­pro­pri­ately] di­rected to change policy could po­ten­tially im­prove a lot of lives, be­cause poli­cies of­ten cover an en­tire coun­try, and for a long time. Changes in policy may last many years into the fu­ture with­out the need for ad­di­tional philan­thropic as­sis­tance.

2. There are some in­ter­ven­tions that can only be done by gov­ern­ments. Thank­fully, NGOs don’t have the power to go into a low-in­come coun­try and just change the law. But that means that if you want to do par­tic­u­lar things, you have to work with gov­ern­ments.

We’re aware that this is go­ing to be a challeng­ing area, both to work in and as­sess. There are a num­ber of rea­sons for that. We think there’s a high chance of failure. I wouldn’t be sur­prised if many of the grants we recom­mend end up hav­ing close to zero effect.

[The first is­sue is that] it’s very hard to pre­dict when [good op­por­tu­ni­ties to set new policy] will open up, and whether the par­tic­u­lar con­text will [al­low us to make] a differ­ence.

The sec­ond is­sue is that causal at­tri­bu­tion to spe­cific ac­tors is par­tic­u­larly hard to de­ter­mine. When a policy change hap­pens, it’s difficult to say, “This group was 20% re­spon­si­ble for it” or “This group was 30% re­spon­si­ble.” We haven’t con­cluded on how to do that. A lot of our rea­son­ing is likely to be more qual­i­ta­tive than it was pre­vi­ously.

The third thing that I think makes policy par­tic­u­larly challeng­ing is it can be difficult — not always, but of­ten — to eval­u­ate a par­tic­u­lar policy change through an ex­per­i­men­tal lens. If a policy is put into place in a coun­try, it will of­ten [af­fect] the whole coun­try at the same time. There won’t be a good treat­ment-con­trol com­par­i­son.

The fourth challenge is some­thing we must be very con­scious of: the risk of ex­ces­sive pa­ter­nal­ism. We need to un­der­stand that, as out­siders, we have limited knowl­edge. We must en­sure that we work through demo­cratic in­sti­tu­tions to the great­est ex­tent pos­si­ble.

That’s a brief up­date about why GiveWell is work­ing in this area. Next, I’ll talk about our ten­ta­tive con­clu­sions so far.

First of all, policy is a huge space. It’s not just one thing, and there are mul­ti­ple ways of carv­ing up the space. What are some things you could do to im­prove how effec­tively gov­ern­ments op­er­ate within a coun­try?

Slide07

One ap­proach might be to help gov­ern­ments choose bet­ter pro­grams. You might want to fund a group to eval­u­ate differ­ent pro­grams and give ad­vice to the gov­ern­ment about which things could be the most effec­tive. If you can im­prove the effec­tive­ness of a gov­ern­ment’s spend­ing, that could be very im­pact­ful.

There are also op­por­tu­ni­ties to in­fluence eco­nomic poli­cies: trade policy, fis­cal policy, in­dus­trial policy, im­mi­gra­tion. High-in­come coun­tries fol­low cer­tain poli­cies that, for poorer coun­tries and poli­ti­cal in­sti­tu­tions, of­ten have nega­tive ex­ter­nal­ities. If we figure out [what those are], I think we will have done very well. Govern­ments don’t just run [pub­lic welfare] pro­grams. Im­prov­ing the tax col­lec­tion sys­tem, the ju­di­ciary sys­tem, how the ad­minis­tra­tion works, and var­i­ous other func­tions of the gov­ern­ment could also be very im­pact­ful.

The area we chose to fo­cus on first is pub­lic health reg­u­la­tion. There are a few rea­sons for that:

1. We think it’s po­ten­tially a par­tic­u­larly im­pact­ful area. For ex­am­ple, if we look back in the past, there are ex­am­ples of to­bacco tax­a­tion [be­ing benefi­cial]. There is fairly com­pel­ling ev­i­dence that, while not ex­per­i­men­tal, in­di­cates that in­creases in to­bacco tax­a­tion gen­er­ally re­sult in de­creased rates of to­bacco-re­lated ill­nesses.

2. We also think there are op­por­tu­ni­ties for out­side philan­thropic as­sis­tance to play a role. It’s not just about fund­ing grass­roots ad­vo­cacy cam­paigns. It’s also about pro­vid­ing tech­ni­cal as­sis­tance and re­search, and there are a lot of ar­eas where out­side as­sis­tance might be [helpful].

3. We wanted to start in a space that [re­flects] where GiveWell is and builds on our ex­pe­rience — a space where we could use the tools and abil­ities that we have built up over time to make good de­ci­sions. What per­haps dis­t­in­guishes pub­lic health reg­u­la­tion from many of the other ar­eas we may have cho­sen is that it is fairly prox­i­mate to im­pact. With poli­ti­cal in­sti­tu­tions, a lot of things must hap­pen be­fore changes filter through and im­prove peo­ple’s lives. Health in­ter­ven­tions are fairly well-defined, so you might be able to take ev­i­dence from one set­ting and ap­ply it to an­other, [scal­ing your im­pact more quickly than you might have been able to in other policy ar­eas].

In our work so far, we have tried to find the ar­eas within pub­lic health reg­u­la­tion that we think GiveWell might want to fo­cus on in the fu­ture.

Slide08

We started with a long list that we took from the Global Bur­den of Disease In­sti­tute and went through differ­ent risk fac­tors, as well as causes of death and dis­abil­ity, and con­sid­ered whether there are reg­u­la­tory op­por­tu­ni­ties that could help to alle­vi­ate this bur­den. Us­ing that heuris­tic, we nar­rowed down the list to nine ar­eas. We also iden­ti­fied a num­ber of other ar­eas that we hope to come back to in the fu­ture.

The pri­mary way that we dis­t­in­guished be­tween those nine ar­eas was to ask: How much money is cur­rently be­ing spent on this? That ques­tion was a proxy for ne­glect. [We also con­sid­ered] the cur­rent dis­ease bur­den of each area.

Once we went through that pro­cess, three ar­eas re­ally stood out:

1. Al­co­hol con­trol;

2. Lead paint elimi­na­tion; and

3. Pes­ti­cide reg­u­la­tion.

I will note that there were a num­ber of difficul­ties and limi­ta­tions of this method­ol­ogy. We hope to go back and re­fine it over time. But I think this is an ex­am­ple of GiveWell’s [con­scious de­ci­sion], when do­ing early work, to take a broad ap­proach rather than do­ing deep re­search in a sin­gle area. We of­ten find that once you’ve [done that type of deep dive], you be­come at­tached to that area. To avoid that, we start broad and shal­low.

Gen­er­ally, if you take how much was spent on those three ar­eas — al­co­hol con­trol, lead paint elimi­na­tion, and pes­ti­cide reg­u­la­tion — di­vide it by the [im­pact of the prob­lem on peo­ple] and then com­pare the re­sult to the other ar­eas we looked at, they were gen­er­ally re­ceiv­ing three to five times less [fund­ing] than some of the other ar­eas we looked at. All three are roughly similar on that met­ric.

Here is a brief overview of al­co­hol con­trol:

Al­co­hol, ac­cord­ing to the Global Bur­den of Disease In­sti­tute, is re­spon­si­ble for about 1.2 mil­lion deaths ev­ery year in low- and lower mid­dle-in­come coun­tries. Our best guess is that un­der $5 mil­lion a year is spent speci­fi­cally on ad­vo­cacy for bet­ter al­co­hol poli­cies. Com­pare that to to­bacco con­trol, which is an area that has re­ceived a lot more at­ten­tion. Tobacco is re­spon­si­ble for about one and a half times more bur­den of dis­ease than al­co­hol, but about 15 times as much is spent on it ev­ery year. So al­co­hol looks to be a par­tic­u­larly promis­ing area. It is su­perfi­cially similar to to­bacco in many ways (al­though there are also im­por­tant differ­ences), but hasn’t re­ceived nearly as much at­ten­tion.

We’re look­ing at what we could do in this area — the best buys. There [are many fac­tors] we need to look into more deeply, such as al­co­hol taxes, mar­ket­ing re­stric­tions, and re­stric­tions on when you can pur­chase al­co­hol.

We have a num­ber of re­main­ing ques­tions in this area. We haven’t fully vet­ted the Global Bur­den of Disease In­sti­tute es­ti­mates, and we also want to be care­ful and con­sider the num­ber of peo­ple who en­joy al­co­hol. How might we fac­tor that into the cost-benefit anal­y­sis?

I’d say what dis­t­in­guishes al­co­hol from the other two ar­eas is the very large bur­den of dis­ease it car­ries. The other two are more “niche” bur­dens, but perform well on the “[fund­ing] ne­glect” crite­rion.

Lead paint elimi­na­tion has been [a big area of in­ter­est] in the U.S. for a while, but I hadn’t heard of it in the U.K., where I grew up. There is some ev­i­dence, con­sis­tent with what we know about lead, that ex­po­sure to lead paint dur­ing child­hood can have large effects on peo­ple’s cog­ni­tive de­vel­op­ment — and that can af­fect their earn­ings later in life. That comes from a se­ries of mul­ti­vari­ate re­gres­sions, which is not com­pletely dis­pos­i­tive, but there are rea­sons to think that this might be true. We know that lead is a toxic metal. We know that a lot of chil­dren in low-in­come coun­tries have high lev­els of lead in their blood. And we know this has to be com­ing from [some­thing in the en­vi­ron­ment], be­cause the nat­u­ral level with­out ex­po­sure to lead is very close to zero, and those coun­tries have, in many cases, already elimi­nated lead in gasoline.

Paint is widely con­sid­ered to be one of the most im­por­tant con­trib­u­tors to [the high lev­els of lead in peo­ple’s blood]. We think un­der $3 mil­lion is spent on lead paint elimi­na­tion, and it ap­pears to be an area that has some trac­tion. A num­ber of coun­tries in the last 20 years have brought in new re­stric­tions.

The third area, pes­ti­cide reg­u­la­tion, is some­thing I talked about at last year’s EA Global con­fer­ence. It’s an area in which we’ve already made a grant. The case for this fo­cus area [rests on] re­duc­ing suicide. About 150,000 peo­ple die ev­ery year from de­liber­ately in­gest­ing pes­ti­cide. There’s some ev­i­dence from Sri Lanka in the 1990s, which used to have one of the high­est suicide rates in the world, that tar­geted re­stric­tions on cer­tain pes­ti­cides co­in­cided with a re­duc­tion in suicides of about 50%. That is one of the largest sus­tained re­duc­tions in suicide in his­tory.

That is liter­ally a time trend, so the ev­i­dence isn’t [ideal]. But it’s backed up by the med­i­cal records. You see a shift in the pes­ti­cides peo­ple drink in suicide at­tempts, from those very high in tox­i­c­ity to those lower in tox­i­c­ity, re­sult­ing in an im­prove­ment in the sur­vival rate. If peo­ple sur­vive a suicide at­tempt, more of­ten than not, they won’t try again.

Th­ese three ar­eas are our cur­rent top pri­ori­ties. We’re definitely go­ing to come back to other ar­eas. There are two I’m par­tic­u­larly ex­cited about. One is iron for­tifi­ca­tion in In­dia, which has a very high bur­den of iron defi­ciency ane­mia. Govern­ments can pass cer­tain reg­u­la­tions to man­date that pro­duc­ers of wheat or rice, for ex­am­ple, put var­i­ous vi­tam­ins into those prod­ucts be­fore they get to the table. We want to do more work on that sce­nario. Another one that we found [promis­ing but] par­tic­u­larly challeng­ing to as­sess is am­bi­ent air pol­lu­tion, which has a very large bur­den of dis­ease. That is tricky, be­cause a lot of the things that you would do to re­duce par­tic­u­late mat­ter — which di­rectly im­pacts hu­man health from air pol­lu­tion — would over­lap sub­stan­tially with what you might do to re­duce CO₂ emis­sions. It was challeng­ing to as­sess within our frame­work, be­cause we had to ask: Do we in­clude all the fund­ing that is spent on CO₂ re­duc­tion?

There are prob­lems with our method­ol­ogy worth not­ing that we hope to go back to and im­prove on in the fu­ture:

Slide09

1. We gen­er­ally use the to­tal bur­den of dis­ease as a proxy for im­por­tance. I think that does leave some­thing out. For ex­am­ple, we’re never go­ing to re­duce al­co­hol con­sump­tion to zero, or at least not in the short term, and it’s not even ob­vi­ous that we should try to. So we might be over­weight­ing some ar­eas just be­cause they have a large bur­den of dis­ease, even if plau­si­bly we could only im­pact them by a small amount, like 10%.

2. It is very challeng­ing to as­sess how tractable par­tic­u­lar policy changes might be. We took this into ac­count some­what [but per­haps not enough]. For ex­am­ple, lead paint elimi­na­tion ap­pears to be an area in which in­dus­try is gen­er­ally on board. That did play some role in our pri­ori­ti­za­tion, but it’s very hard to pre­dict which poli­cies will be tractable at which times, and policy win­dows of­ten open for a very short time. We think we could be miss­ing [im­por­tant fac­tors] if we de­pri­ori­tized an area just be­cause we think it’s hard to gain trac­tion on right now.

3. I think we still have a lot more work to do on the cost-effec­tive­ness of differ­ent reg­u­la­tions. In the fu­ture, we will do that be­fore in­vest­ing in an area.

4. We’re con­sid­er­ing ar­eas at differ­ent lev­els of gran­u­lar­ity. If, in­stead of look­ing at pes­ti­cide suicide speci­fi­cally, we looked at suicide in gen­eral, it’s pos­si­ble we would have come to differ­ent an­swers. In de­pri­ori­tiz­ing some of the ar­eas we ex­plored at a very broad level, we may have [over­looked] spe­cific things within them that are very [worth­while].

What does all of this mean to GiveWell? And what might it mean to you?

Slide11

We’re re­ally ex­cited about this work. We think it’s likely to lead us to be­com­ing far more cost-effec­tive than we cur­rently are, and to recom­mend giv­ing op­por­tu­ni­ties that are far bet­ter than what we cur­rently recom­mend.

We do ex­pect to con­tinue recom­mend­ing top char­i­ties us­ing our tra­di­tional crite­ria in the fu­ture. We don’t think any­thing about that is go­ing to change, at least in the near term, al­though it might well be that if we iden­tify enough policy-re­lated op­por­tu­ni­ties, the amount of fund­ing that goes to our top char­i­ties will be diminished. We still want to keep trans­parency as a core value. We’re still go­ing to com­mit to writ­ing up any recom­men­da­tions we make in great de­tail.

We’re also cog­nizant of the fact that there will be more in­stances of news that will be difficult to share. GiveWell has never re­ally had en­e­mies be­fore. There aren’t groups try­ing to stop bed­nets from be­ing dis­tributed. That might not be the case in ar­eas like al­co­hol or to­bacco, where there are well-funded in­dus­try groups with strong po­si­tions. I think we have to be very care­ful about what we can share and when, while main­tain­ing a com­mit­ment to trans­parency to the ex­tent that we can.

We still ex­pect quan­ti­ta­tive rea­son­ing to play a large role in our work. We still find us­ing spread­sheets very use­ful. But we are go­ing to put less weight on those than we do when we rea­son through di­rect in­ter­ven­tions. And we know that some of the as­sump­tions that [un­der­lie the new work] will be very difficult to jus­tify.

It’s also pos­si­ble that this work won’t lead to any changes. We haven’t com­pleted our in­ves­ti­ga­tions yet and we want to keep an open mind. We don’t want to [take on this new work] just be­cause it’s ex­cit­ing and shiny. My best guess is that we will be able to find op­por­tu­ni­ties in this area; in fact, we have already made two grants. But it’s early days.

Be­fore I finish up, I want to men­tion that I’ve fo­cused on the pub­lic health reg­u­la­tion work we’ve done here. We have also made two grants in policy already. One was to the Cen­tre for Pes­ti­cide Suicide Preven­tion, which I men­tioned. The other was to the In­no­va­tion in Govern­ment Ini­ti­a­tive. That or­ga­ni­za­tion is very ex­cit­ing.

So what does this mean for you?

Slide13

I think two things. One, if you are a donor, I ex­pect that GiveWell will pre­sent a differ­ent suite of giv­ing op­por­tu­ni­ties within the next year or two. We’re un­cer­tain ex­actly how we’re go­ing to do that. It could be [in the form of] some recom­men­da­tions for char­i­ties we think are great. It’s more likely, I think, to be [an ap­proach whereby you could] con­tribute to a fund, and then we would use that to fund op­por­tu­ni­ties that haven’t been defined in ad­vance. With these kinds of op­por­tu­ni­ties, it’s less about ev­ery dol­lar hav­ing a marginal im­pact. Spe­cific grant fund­ing to do spe­cific ac­tivi­ties might be a bet­ter ap­proach.

The other im­pli­ca­tion is that we’re hiring. We’re par­tic­u­larly ex­cited for peo­ple to ap­ply who are in­volved in the effec­tive al­tru­ism com­mu­nity. We’ve had a lot of suc­cess with that in the past. We’re mostly go­ing to be look­ing for ex­pe­rienced re­searchers, man­agers, and peo­ple who have worked in global health and de­vel­op­ment be­fore. But we also ex­pect to be hiring a num­ber of ju­nior staff straight out of col­lege.

Moder­a­tor: The first ques­tion from the au­di­ence: What is your plan for when GiveWell’s ideas or agen­das run afoul of the prefer­ences of poli­ti­cally pow­er­ful or well-funded com­pet­ing in­ter­ests?

James: It’s a good ques­tion. I don’t think I can gen­er­al­ize. Those com­pet­ing in­ter­ests will be differ­ent in each case. So I’m sorry, but I’m not go­ing to be able to give a very satis­fy­ing an­swer to this, ex­cept to say that it’s go­ing to be do­main-spe­cific. It’s go­ing to de­pend on the spe­cific grant, and we think that many of the best peo­ple to be mak­ing those kinds of tac­ti­cal de­ci­sions on the ground [will likely be the peo­ple we’ve funded, rather than our own re­searchers]. So we want to be very hum­ble. We need to rec­og­nize what we’re good at and what we don’t know about, and not micro­man­age grantees.

Moder­a­tor: Do you see op­por­tu­ni­ties in policy in­ter­ven­tion that af­fect mul­ti­ple cause ar­eas si­mul­ta­neously, such as within af­fir­ma­tive vot­ing meth­ods?

James: Yes. I don’t know much about af­fir­ma­tive vot­ing meth­ods in par­tic­u­lar, but I think one of the things that’s challeng­ing is that, [in or­der to make de­ci­sions about what to fund], we carve up the policy space. But that doesn’t re­flect re­al­ity. It’s a very sim­plified model of re­al­ity. So for at least some of these [policy ar­eas], you can think about im­pact in terms of a pyra­mid. Some things are very much at the top and will af­fect mul­ti­ple other things. For ex­am­ple, if you could im­prove poli­ti­cal in­sti­tu­tions in a coun­try, that’s go­ing to have some effect on health even­tu­ally. There are causal ar­eas that work that way.

We de­cided to start, as I said, in an area that is fairly spe­cific and prox­i­mate to im­pact, be­cause we think that’s the most tractable thing to do in the short term. But in the long term, we’d love to look at ar­eas that are much broader.

Moder­a­tor: Last ques­tion: How much pa­ter­nal­ism is ap­pro­pri­ate and not ex­ces­sive?

James: I was afraid some­one might ask that. I think one dis­tinc­tion I’d make is — and I apol­o­gize if I’m butcher­ing the lan­guage here — be­tween what I would call “pa­ter­nal­ism” and what I would call a “neo­colo­nial­ist at­ti­tude.” To me, pa­ter­nal­ism is about in­di­vi­d­u­als and in­di­vi­d­ual re­stric­tions in liberty, whereas a neo­colo­nial­ist at­ti­tude might be closer to not re­spect­ing the demo­cratic norms of a coun­try.

To ad­dress the pa­ter­nal­ism ques­tion, pub­lic health reg­u­la­tion of­ten does have a pa­ter­nal­is­tic el­e­ment. There are some nu­ances around that. For ex­am­ple, with to­bacco taxes, there was strong push­back early on, un­til the sec­ond­hand smoke is­sue was dis­cov­ered. Sud­denly, the fo­cus was on hurt­ing other peo­ple. But I think the biggest gains we’ve had are health benefits for the in­di­vi­d­u­als who smoke.

It’s very difficult to com­pletely di­vorce this from one’s poli­tics. If you are a hard­core liber­tar­ian, I don’t think you should give to these par­tic­u­lar [policy] op­por­tu­ni­ties. But I think there are some cases in which peo­ple are sys­tem­at­i­cally mak­ing bad de­ci­sions, and that’s not just be­cause they live in low-in­come coun­tries. That’s true in high-in­come coun­tries as well. So if we feel fairly com­fortable with be­ing some­what pa­ter­nal­is­tic in the U.S. or U.K., I don’t think that there’s nec­es­sar­ily a rea­son to be much more cau­tious in other coun­tries, un­less there are par­tic­u­larly differ­ent so­cial norms in those places.

No comments.