The Measuring people’s preferences IDinsight survey shows that the average life satisfaction score of Kenyan respondents is 2.3/10 (n = 1,808, SD = 2.32) (p. 42). If the neutral point, “the point where someone is neither satisfied nor dissatisfied,” is above 2.3/10, reducing mortality in Kenya by malaria vaccination can save lives of dissatisfaction.
This also seems inaccurate/misleading. From Page 40 of the report:
On the ladder from 0 to 10, respondents (n = 1808) report an average life satisfaction score of 2.8 (range 0 − 10, standard deviation = 2.32). This score is much lower than the 2019 World Happiness Report – Kenya had a score of 4.66 while Ghana had a score of 5.48.68 This survey uses nationally representative samples, which means the average respondent would be richer than our purposefully sampled respondents from low-income regions and households. However, our results are still lower than expected based on extrapolation from other studies (see Appendix 6 for further exploration of this pattern).
There are two issues with your summary, one minor and one serious:
The more minor error is that you reported 2.3 as 2.8 (maybe because you copied the s.d.?)
The more major issue is that you reported the numbers from the IDInsight survey at face value, without noting that it was written in the overall context of being an outlier from other surveys. This may lead your readers to systematically underestimate life satisfaction numbers in Kenya.
That said, I did not consciously realize that average life satisfaction numbers are so low in Kenya/Ghana. This is helpful context for me, and makes the value of poverty alleviation efforts more visceral.
I looked at the graph on page 42 (the bar for Kenya is 2.3), but actually had the statistic from HLI, which cites it. The 2.8 (p. 40) is the survey average. Good catch for HLI (and myself, unless I am further misreading), which can make it seem as if the Kenyan sample was 1,808 and SD=2.32.
OK, I was actually unaware of that (clearly was skimming the HLI page for bias confirmation—or, rather, made a note of an alarming statistic when skimming). I added the 4.5 value from the 2019 World Happiness Report also cited by HLI. This averages closer to 3.9/10, which is the LS estimate for GiveDirectly beneficiaries.
Thanks that makes sense! I did not realize that the average in Kenya for the IDInsight surveyed sample is 2.3. I appreciate my correction being corrected.
The IDInsight n from poor people surveyed in Kenya is 954 in case this is relevant to you. Appendix 1, Page 58 in the report.
I added the 4.5 value from the 2019 World Happiness Report also cited by HLI.
I was a bit confused about how the report said 4.5 for Kenya while IDInsight said 4.66 for WHR. My current guess was that 4.5 was for “happiness” while 4.66 was for “life satisfaction.” However I could not find life satisfaction numbers in Kenya on a quick skim in WHR, will encourage other data sleuths to pick up the slack if desirable!
If we take both the 2.3 and 4.5 numbers at face value (and I’m not sure we should, but I don’t have strong ideas otherwise for how to make donation decisions in the near future otherwise), one plausible interpretation is that average Kenyans are significantly happier than the poorest Kenyans. This will all else equal favor consumption and growth interventions over lifesaving measures (though of course there are many other considerations in place).
This will all else equal favor consumption and growth interventions over lifesaving measures (though of course there are many other considerations in place).
Yup, assuming causality.
[D]oubling consumption corresponds to a 0.42 increase in the life satisfaction score … Our ‘wealthy’ households had an average life satisfaction score of 4.3, while the ‘poor’ households had an average life satisfaction of 2.8. (p. 42) … Stevenson and Wolfers (2013) finds a lower coefficient of 0.25 among lower income countries (p. 41)
I would be careful about simply increasing consumption and growth. More marketing (including that which highlights negative/abusive cultural aspects) could enter areas where identities are otherwise based in emotional navigation of relationships, which can be understood as deeply satisfying (these identities would be lost with increased societal attention paid to current globally competitive marketing).
Perhaps, this would start from an income level that would not be reached even with income doubled a few times, but, considering very affordable products, the Belt and Road Initiative, and growing marketing analysis and capacity in rapidly growing countries in Asia, growth without co-interventions can lead to an increased consumption of ‘aggressively’ marketed products, which may not increase one’s life satisfaction.
This paper on cultural combination (‘syncretism’) from the South African University of Pretoria. There is little on the possibility of ‘disturbing’ pictures or arguably sexist bias-based and objectifying/physically judging advertisements becoming popular among some people. It is unlikely that the people affected by the marketing (even non-customers) would be interacting with humans of different cultures (but rather see the ads which do not respond to human emotional expressions).
People could be reporting an ‘objective’ life satisfaction, based on status portrayed in the ads, without emotional introspection. It is possible that they would not report dissatisfaction, because that would mean decreased competitiveness, which, based on some advertisements, could be associated with one’s vulnerability or undesirable situation/identity. This is just a hypothesis.
Also, the lives of the poorer persons can be worse because of the norms that they grow up in (for example, threatening of neighbor’s life for $3, sending children to work or beg from a very young age, defaulting on a group loan, … vs. going to different neighbors for humble meals weekly, trying to put children through school, vetting microfinance firms and contemplating the EV of an income-generating asset lease).
The argument is that if you increase the (for instance) children’s who grew up begging income, it does little for them because of their upbringing (it may be difficult for them to form enjoyable relationships because they are used to a lot of unwelcomingness). A better approach would be education in locally relevant skills so that they can be (considering the situation) welcome since a young age.
An alternative thinking is that the people who had limited opportunities when they were young would be super grateful for the improved opportunities and will educate their children so that they do not experience low life quality rather than approaching them as people would approach a begging child (illustrative example of gratitude of situation improvement—actually life saved—from an island I’ve seen). This suggests that the present adult generation should be targeted with consumption increase programs rather than children educated. Saving lives, at least by caring individuals sincerely interested in the saved people, can be actually also valued.
Still, at least some budget should probably be allocated to the “other considerations,” just to make sure that it is not that, for example, men who beat their wives and women who would perpetuate the normalization of beating are not just going to get more colorful washing baskets with ‘women overpowering men by using the product’ for the women. I argued similarly here.
This is a bit of a digression, but I would generally recommend against a) linking the original source directly if you’ve only read the summary (instead you should probably cite/link the summary first), and b) (although this is more work) generally trusting summaries for data/figures without doing quick epistemic spot checks on on the quality of the summaries.
a) While in formal writing, there are specific formats of citing others’ citations, in this context, I decided to link the report directly, alongside with this comment thread that reads
I added the 4.5 value from the 2019 World Happiness Report also cited by HLI.
In this comment, the HLI’s Estimating moral weights page (with the footnote) to which I referred several times in this thread is not referenced, because I assumed that those who read this thread carefully are already familiar with the page and those who are quickly skimming do not need to be distracted by that link.
I am keeping in mind that this is the Change Our Mind contest. Citing HLI could be read as an intent to convince GiveWell to implement HLI’s framework, which they are familiar with, by repetition. WHR allows readers to form and update their opinions based on data which does not intend to change GiveWell’s mind. Thus, WHR can change the mind of an evidence-based decisionmaker better.
Further, historically, GiveWell has used top statistical evidence to make its recommendations. WHR enjoys similar level of comprehensiveness as RCT-based research, while HLI’s research is more speculative. Thus, WHR can allow GiveWell to change their mind more consistently with its fundamental values than HLI’s research.
b) I have not checked the Report, but rather deferred to HLI’s standards of citing statistics. I reviewed some papers cited by HLI and did not find inconsistency (other than the vague sample size interpretation as further above in this thread). This can be understood as a form of a spot check.
Nevertheless, I searched for the statistic in the 2019 WHR. (I used the search function for “4.5” and “Kenya”.) “Kenya (4.509)” is cited as the value on p. 29 of the WHR pdf (pp. 26–27 of the document). I added the page reference.
This actually leads me to the methodology of the WHR. It seems like ‘happiness’ is a function of (pp. 26–27):
GDP per capita
Social support
Healthy life expectancy
Freedom to make life choices
Generosity
Perceptions of corruption
(Constant)
Although this can cover many aspects of happiness, other factors which could influence this metric (including by changing its sign), such as the normality of abuse or parental acceptance/rejection, do not seem to be included. WHR ‘happiness’ can thus measure governance quality and public cooperation rather than seek to understand intended beneficiaries’ quality of life. However, further research is needed.
I also added a note on the interpretation of this metric.
This also seems inaccurate/misleading. From Page 40 of the report:
There are two issues with your summary, one minor and one serious:
The more minor error is that you reported 2.3 as 2.8 (maybe because you copied the s.d.?)
The more major issue is that you reported the numbers from the IDInsight survey at face value, without noting that it was written in the overall context of being an outlier from other surveys. This may lead your readers to systematically underestimate life satisfaction numbers in Kenya.
That said, I did not consciously realize that average life satisfaction numbers are so low in Kenya/Ghana. This is helpful context for me, and makes the value of poverty alleviation efforts more visceral.
I looked at the graph on page 42 (the bar for Kenya is 2.3), but actually had the statistic from HLI, which cites it. The 2.8 (p. 40) is the survey average. Good catch for HLI (and myself, unless I am further misreading), which can make it seem as if the Kenyan sample was 1,808 and SD=2.32.
OK, I was actually unaware of that (clearly was skimming the HLI page for bias confirmation—or, rather, made a note of an alarming statistic when skimming). I added the 4.5 value from the 2019 World Happiness Report also cited by HLI. This averages closer to 3.9/10, which is the LS estimate for GiveDirectly beneficiaries.
Thanks that makes sense! I did not realize that the average in Kenya for the IDInsight surveyed sample is 2.3. I appreciate my correction being corrected.
The IDInsight n from poor people surveyed in Kenya is 954 in case this is relevant to you. Appendix 1, Page 58 in the report.
woohoo thanks.
I was a bit confused about how the report said 4.5 for Kenya while IDInsight said 4.66 for WHR. My current guess was that 4.5 was for “happiness” while 4.66 was for “life satisfaction.” However I could not find life satisfaction numbers in Kenya on a quick skim in WHR, will encourage other data sleuths to pick up the slack if desirable!
If we take both the 2.3 and 4.5 numbers at face value (and I’m not sure we should, but I don’t have strong ideas otherwise for how to make donation decisions in the near future otherwise), one plausible interpretation is that average Kenyans are significantly happier than the poorest Kenyans. This will all else equal favor consumption and growth interventions over lifesaving measures (though of course there are many other considerations in place).
Yup, assuming causality.
I would be careful about simply increasing consumption and growth. More marketing (including that which highlights negative/abusive cultural aspects) could enter areas where identities are otherwise based in emotional navigation of relationships, which can be understood as deeply satisfying (these identities would be lost with increased societal attention paid to current globally competitive marketing).
Perhaps, this would start from an income level that would not be reached even with income doubled a few times, but, considering very affordable products, the Belt and Road Initiative, and growing marketing analysis and capacity in rapidly growing countries in Asia, growth without co-interventions can lead to an increased consumption of ‘aggressively’ marketed products, which may not increase one’s life satisfaction.
This paper on cultural combination (‘syncretism’) from the South African University of Pretoria. There is little on the possibility of ‘disturbing’ pictures or arguably sexist bias-based and objectifying/physically judging advertisements becoming popular among some people. It is unlikely that the people affected by the marketing (even non-customers) would be interacting with humans of different cultures (but rather see the ads which do not respond to human emotional expressions).
People could be reporting an ‘objective’ life satisfaction, based on status portrayed in the ads, without emotional introspection. It is possible that they would not report dissatisfaction, because that would mean decreased competitiveness, which, based on some advertisements, could be associated with one’s vulnerability or undesirable situation/identity. This is just a hypothesis.
Also, the lives of the poorer persons can be worse because of the norms that they grow up in (for example, threatening of neighbor’s life for $3, sending children to work or beg from a very young age, defaulting on a group loan, … vs. going to different neighbors for humble meals weekly, trying to put children through school, vetting microfinance firms and contemplating the EV of an income-generating asset lease).
The argument is that if you increase the (for instance) children’s who grew up begging income, it does little for them because of their upbringing (it may be difficult for them to form enjoyable relationships because they are used to a lot of unwelcomingness). A better approach would be education in locally relevant skills so that they can be (considering the situation) welcome since a young age.
An alternative thinking is that the people who had limited opportunities when they were young would be super grateful for the improved opportunities and will educate their children so that they do not experience low life quality rather than approaching them as people would approach a begging child (illustrative example of gratitude of situation improvement—actually life saved—from an island I’ve seen). This suggests that the present adult generation should be targeted with consumption increase programs rather than children educated. Saving lives, at least by caring individuals sincerely interested in the saved people, can be actually also valued.
Still, at least some budget should probably be allocated to the “other considerations,” just to make sure that it is not that, for example, men who beat their wives and women who would perpetuate the normalization of beating are not just going to get more colorful washing baskets with ‘women overpowering men by using the product’ for the women. I argued similarly here.
The 4.5 is footnote 30 in the HLI summary.
This is a bit of a digression, but I would generally recommend against
a) linking the original source directly if you’ve only read the summary (instead you should probably cite/link the summary first), and
b) (although this is more work) generally trusting summaries for data/figures without doing quick epistemic spot checks on on the quality of the summaries.
a) While in formal writing, there are specific formats of citing others’ citations, in this context, I decided to link the report directly, alongside with this comment thread that reads
In this comment, the HLI’s Estimating moral weights page (with the footnote) to which I referred several times in this thread is not referenced, because I assumed that those who read this thread carefully are already familiar with the page and those who are quickly skimming do not need to be distracted by that link.
I am keeping in mind that this is the Change Our Mind contest. Citing HLI could be read as an intent to convince GiveWell to implement HLI’s framework, which they are familiar with, by repetition. WHR allows readers to form and update their opinions based on data which does not intend to change GiveWell’s mind. Thus, WHR can change the mind of an evidence-based decisionmaker better.
Further, historically, GiveWell has used top statistical evidence to make its recommendations. WHR enjoys similar level of comprehensiveness as RCT-based research, while HLI’s research is more speculative. Thus, WHR can allow GiveWell to change their mind more consistently with its fundamental values than HLI’s research.
b) I have not checked the Report, but rather deferred to HLI’s standards of citing statistics. I reviewed some papers cited by HLI and did not find inconsistency (other than the vague sample size interpretation as further above in this thread). This can be understood as a form of a spot check.
Nevertheless, I searched for the statistic in the 2019 WHR. (I used the search function for “4.5” and “Kenya”.) “Kenya (4.509)” is cited as the value on p. 29 of the WHR pdf (pp. 26–27 of the document). I added the page reference.
This actually leads me to the methodology of the WHR. It seems like ‘happiness’ is a function of (pp. 26–27):
GDP per capita
Social support
Healthy life expectancy
Freedom to make life choices
Generosity
Perceptions of corruption
(Constant)
Although this can cover many aspects of happiness, other factors which could influence this metric (including by changing its sign), such as the normality of abuse or parental acceptance/rejection, do not seem to be included. WHR ‘happiness’ can thus measure governance quality and public cooperation rather than seek to understand intended beneficiaries’ quality of life. However, further research is needed.
I also added a note on the interpretation of this metric.