I was pretty taken aback by GiveWell’s moral weights by age. I had not expected them to give babies such little moral weight compared with DALYs. This means GiveWell considers saving babies’ lives to be only as valuable as saving people in their late 30s despite them being almost halfway through their life. The graph makes the drop-off of moral weights at younger ages look less sharp than it is as the x-axis is not to scale.
I looked at the links for further information on this which I’m collating here for anyone else interested:
These results look sensible to us. We’re least certain about the value of averting deaths at very young ages and stillbirths. If those values became decision-relevant for a grant, such as a neonatal health program, we might revisit these weights or consider setting aside a pot of funding for grants that satisfy “other reasonable moral weights.”
DALYs assume that preventing the deaths of people with longer remaining life expectancies is always more valuable, when in practice many people indicate a preference for preventing the death of an older child over the death of a neonate.[footnote]
The footnote:
This applies to GiveWell staff and donors, as well as to the results of the Mechanical Turk survey referenced here. The IDinsight survey did not ask about neonates.
The biggest uncertainty we have is around the relative value of preventing deaths at very young ages.
This is in context of the ID survey but I assume is speaking about GiveWell’s uncertainty not the uncertainy of ID survey respondents.
We don’t believe the group of donors we surveyed is very diverse (across characteristics like race, gender, income, and country of origin) which could influence results. The vast majority of the donors we surveyed are men, and people of different genders could especially have different intuitions about the value of averting stillbirths and the deaths of neonates.
A quick analysis of the responses of men vs. women didn’t indicate that we should upweight stillbirths and the deaths of neonates to account for different preferences across genders, but there were so few women in the sample that we can’t say with confidence that the results don’t depend on gender.
Apparently the moral weights have not been decision-relevant which is good news for all donors who have different preferences for moral weights. In the future I will check before donating to GiveWell’s funds whether the moral weights of babies have become decision-relevant for grants in the meantime.
I was also wondering how many of the donor survey respondents were parents and whether they put moral weight on babies than non-parents. I could also imagine there being a discrepancy between mothers and everyone else (fathers and the childless).
I was pretty taken aback by GiveWell’s moral weights by age. I had not expected them to give babies such little moral weight compared with DALYs. This means GiveWell considers saving babies’ lives to be only as valuable as saving people in their late 30s despite them being almost halfway through their life. The graph makes the drop-off of moral weights at younger ages look less sharp than it is as the x-axis is not to scale.
I looked at the links for further information on this which I’m collating here for anyone else interested:
From the [public] 2020 update on GiveWell’s moral weights—Google Docs:
The footnote:
This is in context of the ID survey but I assume is speaking about GiveWell’s uncertainty not the uncertainy of ID survey respondents.
Apparently the moral weights have not been decision-relevant which is good news for all donors who have different preferences for moral weights. In the future I will check before donating to GiveWell’s funds whether the moral weights of babies have become decision-relevant for grants in the meantime.
I was also wondering how many of the donor survey respondents were parents and whether they put moral weight on babies than non-parents. I could also imagine there being a discrepancy between mothers and everyone else (fathers and the childless).