I think you have two issues in your discussion of the differential effects of malaria prevention on men’s and women’s incomes.
I think you are misinterpreting GiveWell’s adjustment for this effect when you say “we take this to imply that GiveWell thinks that the effect on women’s income is 25% lower than that for men’s.” Looking in detail at the calculation behind GiveWell’s replicability adjustment, it seems clear to me that they are multiplying the overall income gain by 0.75, not just women’s. Therefore, they are assuming that women get 50% of the income boost that men get, not 75%
I think you have a math error in your updated calculation. You compute that if women get 21% of the income boost that men do, it would imply a ~39% downward adjustment (cell C11). But you implement that by multiplying GiveWell’s replicability adjustment directly by your downward adjustment (cell C110). The effect of this is to apply a ~61% decrease in the benefits, rather than the intended ~39% decrease. I think the right way to implement this is to multiply the replicability adjustment by one minus your downward adjustment. Correcting this error, I estimate that this change leads to a 6% decrease in AMF’s cost-effectiveness, rather than the 16% decrease you initially reported.
GiveWell’s description of their adjustment:
“The studies primarily report outcomes for males. We would guess that there would also be effects on income for females, but that these would be lower due to females having lower rates of labor force participation. We apply a further discount of 75% to account for this consideration. 70% x 75% gives a replicability adjustment of 52%.”
I think you have two issues in your discussion of the differential effects of malaria prevention on men’s and women’s incomes.
I think you are misinterpreting GiveWell’s adjustment for this effect when you say “we take this to imply that GiveWell thinks that the effect on women’s income is 25% lower than that for men’s.” Looking in detail at the calculation behind GiveWell’s replicability adjustment, it seems clear to me that they are multiplying the overall income gain by 0.75, not just women’s. Therefore, they are assuming that women get 50% of the income boost that men get, not 75%
I think you have a math error in your updated calculation. You compute that if women get 21% of the income boost that men do, it would imply a ~39% downward adjustment (cell C11). But you implement that by multiplying GiveWell’s replicability adjustment directly by your downward adjustment (cell C110). The effect of this is to apply a ~61% decrease in the benefits, rather than the intended ~39% decrease. I think the right way to implement this is to multiply the replicability adjustment by one minus your downward adjustment. Correcting this error, I estimate that this change leads to a 6% decrease in AMF’s cost-effectiveness, rather than the 16% decrease you initially reported.
GiveWell’s description of their adjustment:
Hi MHR,
Thanks for this! You’re definitely right about the text, and I double checked the numbers and also think you’re right there.
Thanks for checking!