3 ms·

The mention of cigarette spending is an extremely dangerous strawman, largely because for some reason people actually believe it. There's fairly strong evidence

• notahacker 10y ago
  Appreciate the sentiment, but it was actually a quite deliberate comparison because (i) the very poor in developing countries where cigarettes are very cheap often have a cigarette budget which comes before investment and improved nutrition (ii) I'm well aware that despite the startling prevalence of smoking amongst poor men, increased tobacco use isn't a significant factor in studies of how poor people spend donor cash. (i) pretty much underlines my point about the things people choose to spend their money on not necessarily being what's best for them. And in the short term it's not an inexplicable phenomenon because cheap tobacco probably does provide more pain and stress relief than slightly better nutrition. (ii) is partly because surveys like those you've posted accurately represent people tending not alter their habits of smoking cheap cigarettes (or frowning on tobacco) after receiving a sudden windfall; there's a limit to how much one can smoke and stockpiling makes no sense when for people who don't even have a waterproof roof. But it's also because poor people are smart enough to realise that people quizzing them on their lifestyle and unexpected windfalls from foreign donors aren't unrelated, and the natural response is to tailor responses to what they think the foreign donors want to hear, resulting in any increases in spending on cigarettes and alcohol (whilst likely low relative to increases in spending on roofs or microbusiness investments anyway) being systematically underreported.
  • glomph 10y ago
    Do you have any evidence for systematic underreporting? One could just as easily say that the members of control groups (who do not receive cash) would have a reason to underreport their spending on things that are not approved of in order to receive more potential money in the future. Also the research presented by GiveDirectly also includes studies of people where the cash is not from the same people as the surveys (for example state funded benefit programs that have clear and universal criteria unrelated to spending or consumption.) In those cases again there is reason to expect the underreporting to be similar across both groups that get cash and those that do not currently.
    • notahacker 10y ago
      > One could just as easily say that the members of control groups (who do not receive cash) would have a reason to underreport their spending on things that are not approved of in order to receive more potential money in the future This apparently happening with control groups in the same villages as cash transfer recipients is actually some of the most persuasive evidence response biases exist in the surveys. According to various RCTs, when their neighbours start getting money, women are less likely to report their husbands beating them. On the other hand, they're also less likely to report being satisfied with their life. There are [just about] plausible mechanisms for how spillover effects from donations could cause those behavioural/satisfaction changes amongst non-recipients in a neighbourhood, but recipients thought they were less likely to receive money if they reported domestic violence or said they were satisfied with their life is the most Ockham's-razor satisfying one. (It's also likely part of the reason why recipients seem surprisingly unlikely to report increased satisfaction from life after receiving life-changing sums despite positive changes in almost every measurable area.) Response biases are difficult enough to eliminate in social science surveys when participants have no reason to suspect a sum greater than their average annual income may hinge on their answers.