Thanks for the comment! Could you clarify why you think paying less is fine, but having lower safety standards is not? Increasing safety standards increases the cost of hiring workers, and therefore increases unemployment. Assuming a prospective worker if aware of the risks of the job, and still thinks the job is worth it, would you oppose the worker taking the job?

Thanks for sharing! Strongly upvoted. I like the study randomised people to receiving job offers instead of being placed into jobs. The opportunity to reject the offer makes it more like the real world.

Here is how the authors describe the effects on income.

On balance the factory job offer seems to have no significant effect on income by any of the three measures, and a family index of the three increases only 0.014 standard deviations.37 These income measures tend to be skewed and highly variable, however, and so our estimates are imprecise. Thus while the average effect is close to zero, the confidence interval on income includes moderate increases and decreases in income from the industrial job offer.

I guess the effect was slightly positive, but that the study was underpowered to detect it. The authors also note the workers think the increased health risks are worth it.

Naturally, workers were probably not perfectly informed about job risks and quality of these jobs, and there is some evidence that they underestimated the risk. Nonetheless, our data suggest that workers understood the health risks, at least in part, did not update their assessment of the risks as a result of spending more time in industry, and that they were willing to bear these risks to cope with temporary unemployment spells.

workers who got jobs in sweatshops were not necessarily better off than those who did

Nitpick. There is a "not" missing at the end.

Thanks for the good point! The 1st edition has 198 pages, and the 2nd has 236, so only 19.2 % (= 236/198 - 1) more pages.

I agree workers have incomplete information, but it is unclear to me why they would systematically overestimate the value of jobs. If it is possible for foreign consumers to know that products from a given company or country were produced chaining workers to machines, I would expect the workers to also know about it given they have a much greater incentive to figure out their working conditions, as they directly reflect on their quality of life. My argument does assume workers freely take up jobs. The assumption that workers can freely leave their jobs is not strictly necessary, but it does help. In general, the greater the freedom of the workers to take up and leave jobs, the greater the transparency about the working conditions, and the greater the decision capacity of workers (adults have more than children), the stronger the case for assuming that taking the jobs was for the best, which I think is in agreement with what you are saying.

Thanks for the comment, Ian. I think the conclusions hold as long as workers are calibrated about the distribution of their working conditions, including their pay, workload, treatment by managers, freedom to leave the job, go on strike, and coordinate with other workers, among others. A sufficiently low chance of being chained to machines is better than starving, and this may be the realistic alternative in some cases. So I think the existence of some of the working conditions you mentioned by itself is not enough to conclude it would be better for them not to exist. One would have to believe the workers are overestimating the quality of their working conditions (in expectation; it is always the case that some workers overestimate and others underestimate the quality of their working conditions).

Hi Ian,

I did not crosspost this due to new supporting quantitative evidence, but I think this is presented in the 2nd edition of Benjamin's book Out of Poverty: Sweatshops in the Global Economy (Cambridge Studies in Economics, Choice, and Society), which was published in January 2025.

Hi artilugiu,

You may be interested in Karthik Tadepalli's series on economic growth in low and middle income countries (LMICs). Part 3 addresses labour markets.

My recommendation for helping people in LMICs is donating to the High Impact Philanthropy Fund (HIPF) from the Centre for Exploratory Altruism Research (CEARCH). I estimate it is 12 times as cost-effective as GiveWell's top charities from the mean between the lower and upper bound of 9 and 15 mentioned by Joel Tan, CEARCH’s founder and managing director, on 28 May 2025. CEARCH estimated the cost-effectiveness accounting only for effects on humans, as a fraction of that of GiveWell’s top charities, of donating to Giving What We Can (GWWC) in 2025 to be 13, that of “advocacy for top sodium control policies to control hypertension” to be 31, that of advocating for “increasing the degree to which governments respond with effective food distribution measures, continued trade, and adaptations to the agricultural sector” in “global agricultural crises [such as nuclear and volcanic winters]” to be 33 (although I estimated this should be 12.4 % as high), and that of “advocacy for sugar-sweetened beverages [SSBs] taxes to control diabetes mellitus type 2” to be 55.

Thanks for the post, Emma and Lewis!

This is a big deal. Because this progress has compounded over multiple generations of hens, well over 300 million individual birds have now been spared life in a battery cage. And, assuming this progress sticks, it won’t be long before that number is a billion. I think this will be the most animals any animal welfare intervention has ever helped.

The Shrimp Welfare Project (SWP) estimates the electrical stunners they bought will be helping 3.3 billion shrimp per year once they are all operational, 11.0 (= 3.3*10^9/(300*10^6)) times the 300 M hens helped by cage-free corporate campaigns. In terms of welfare, for my estimate of 0.0426 DALYs averted per shrimp helped by SWP, this has averted 141 MDALY (= 0.0426*3.3*10^9). For my estimates of 0.452 DALYs averted per hen-year improved by cage-free corporate campaigns, and 1.34 hen-years per hen, the campaigns avert 0.606 DALYs per hen (= 0.452*1.34), and have therefore averted 182 MDALY (= 0.606*300*10^6) in total, 1.29 (= 182/141) times as much as SWP has. There is lots of uncertainty in many numbers. I would say SWP over the past few years has increased the welfare of shrimps roughly as much as all cage-free corporate campaigns have increased the welfare of hens.

Accounting for soil nematodes, mites, and springtails, I think GiveWell's top charities have increased animal welfare much more than cage-free corporate campaigns. I estimated GiveWell's top charities are 8.05 (= 1.69/0.210) times as cost-effective as cage-free corporate campaigns accounting for target beneficiaries and soil animals, and I also think GiveWell's top charities have received much more funding, such that their cumulative impact has been much more than that of cage-free corporate campaigns.

Then advocates got to work. With fewer than a thousand full-time advocates globally, the movement waged campaigns that have now spared over 300 million hens from cages. That’s over 300,000 hens per advocate.

Interesting! Assuming the number of advocates increased from 0 to 1 k over 13 years^[1], 2 k hours per advocate-year, an hourly rate of 25 $/h, and 10 hen-years improved per hen helped due to cage-free corporate campaigns anticipating the transition to cage-free by 10 years, the cost-effectiveness would be 9.23 hen-years improved per $ (= 300*10^6*10/((0 + 1)/2*10^3*13*2*10^3*25)), which is 85.5 % (= 9.23/10.8) of what I assumed in my analysis.

1. ^{^}

   But take a moment to reflect how far we’ve come. A mere 13 years ago, the Humane Society of the US, despairing of ever getting the US egg industry to go cage-free, backed a compromise to legislate an industry transition to larger “enriched cages.” When the pork industry — triggered by the specter of any farm animal welfare legislation — sank the compromise, it looked like battery cages were here for good.

Hi Mata,

The 2nd edition of Benjamin's book I linked at the start of the post was published in January 2025. I have now added this information to the start of the post.

I have not put much effort into finding critiques. I already had the prior that people working in sweatshops is good relative to the counterfactual. Otherwise, I would not have expected them to freely take the job. As a rule of thumb, I defer to people about what is best for them. This is not because I think people make optimal career choices. It is because I am sceptical of my ability to do better with very little knowledge about their situation (for example, just their country).

Thanks for the reply.

Could you elaborate on which type of uncertainty makes you discount effects on wild animals? I assume you are not neglecting these just because they have a high chance of being negligible. This also applies to interventions helping farmed insects, and you have made a grant to the Insect Welfare Research Society (IWRS). Rethink Priorities' (RP's) estimate for the probability of sentience of silkworms is just 1.21 (= 0.082/0.068) times their estimate for nematodes.

I guess you are neglecting effects on wild animals because the probability of them being beneficial is similar to that of them being harmful. Do you have any thoughts on the reasons I presented for that not being a sufficient reason to neglect the effects?