Thankyou to Isaac King and Scott Alexander for help in gathering respondents.
As some of you know, I recently sent out a survey.
For a while now, I’ve been pursuing a theory that impartial altruism explains a portion of the left-right ideological gap. As a causal hypothesis, this is tricky to test, so instead I’ve contented myself with the more modest theory that impartial altruism strongly correlates with being leftwing.
I’ve tested the relationship in the past with self-report questions about attitudes- would you let your friend die to save a Nelson Mandela or Norman Borlaug figure, but this opens up the objection ‘what if the leftwingers are just virtue signaling’? I’ve tested the relationship by looking at altruistic kidney donation- but altruistic kidney donors are a special breed and may not be indicative of larger trends. I’ve tested the relationship between altruism and the left by looking at donations- but religiosity explains most of the variation in yearly donations. What was needed then was a de novo test of altruism.
I decided to set up a survey with an attached lottery. The winner would receive $50 . While filling out the survey, the participant had to describe how they would split the money between themselves and a charity of their choice- with the sweetener that the money they donated would be tripled. The amount the participant chooses to give away would be taken as an indicator of altruism.
Left versus right on generosity
I had participants rate themselves on a 0-10 political scale- 0 being wholly leftwing, 10 being wholly rightwing. Of those who gave answers 0-4 (which I equated with leftism) and of those who gave answers 6-10 (which I equated with rightism), who was more likely to donate a large portion of the money? I won’t keep you in suspense, it was the left.
Why treat left and right as categorical variables? In essence, because I don’t trust that the scale is linear. I suspect that a 4 is much closer to a 0 than a 6. Similarly, I converted giving into a categorical variable, because it was overwhelmingly centered on people who chose to give all of the money away or keep it all. Hence I coded everyone who gave 25 dollars (out of 50) or more to charity as ‘generous’ and those who gave less as ungenerous. This is not a value judgment, just a label in this specific context- I’m sure many of the non-givers had excellent reasons. Anyway, if anyone wants to rerun the analysis as an ordinal regression, let me know.
When I did the test (results are still coming in) here is what I found. Among rightists, 55/165 were ungenerous. Among leftists, 67/419 were ungenerous. This gives an odds ratio of 2.8 to 1. P < 0.0001- that is to say there is less than a 1 in a million chance of getting samples this far from equality if the null hypothesis that the two populations have the same proportions is true.
In other words, assuming no coding errors, there is little room for the result to be a fluke. You are almost three times more likely to give less than 25$ out of 50 if you are rightwing. Although this is a large effect, and likely to be of vital practical significance if it is reflected in the population as a whole, it is worth keeping in mind that this still leaves two-thirds of rightists giving at least 25 dollars. Indeed, 57% of rightwingers gave all of the money, (compared to 75% of leftwingers).
As always, imperfections in the measures of the two correlates also likely dampen the result. A single-question measure of left-right orientation leaves much to be desired, and generosity also cannot be reliably assessed with a single data point. Without these measurement unreliabilities, the real relationship is likely to be stronger.
The two big remaining questions then are:
Is the measure of generosity an adequate measure?
Is the sample an adequate reflection of the population
The purpose of the measure was the following. Rightwing people, being much more likely to go to church, tend to give in church because it’s strongly expected that you will do so - they pass a plate and everything! They are also often expected to tithe their income, in addition to somewhat stochastic donations. It can even become highly socially expected to give to religious charities outside the church, but closely connected with it. But giving to churches and their closely related charities is quite sociologically and psychologically different from other forms of altruism.
Yet, for obvious reasons, we can’t just look solely at giving to secular charities and consider that a ‘fair’ measure of altruism- it would be deeply unfair to religious respondents.
One easy solution would be to control for religiosity, and I support studies that use this approach, but in some ways, they are not ideal. Comparing like amounts of religiosity to like amounts of religiosity might be deceptive, because leftwing Christianity and rightwing Christianity, and leftwing atheism and rightwing atheism, might be different in their social contexts and ideological dimensions. E.g. different levels of expected giving in rightwing versus leftwing churches, different levels of religiosity masquerading under similar answers and so on and so forth.
What was needed then, in my view, was to test giving in a shared context, hence the questionnaire. I think, at its face, the measure is an adequate measure of de novo generosity, although for reasons I discuss in the overall conclusions, it may not be a test of impartial generosity, but generosity in a broader sense (inclusive of generosity to those who are not family or friends, but are still part of the in-group in some sense).
I have significant concerns about the population sampled, to be honest. I suspect the result would replicate in the real population, but who can say? It seems possible that the leftists who read Scott’s blog (where a lot of the respondents came from) are unusually generous, and the rightists who read Scott’s blog are unusually generous. I doubt this is true- certainly, I very strongly doubt it is true with the strength to undo or reverse the result, but this is only my opinion.
Our measure of generosity is close to dichotomous. This leaves the possibility of multiple interpretations in the following sense: We do not know whether rightwingers in general are less generous, or whether there is simply a large subpopulation of rightwingers who are extremely ungenerous and most rightwingers are just as generous or almost as generous as the typical leftwinger. In other words, is the mean shifted, or is there a strong skew? In order to assess this, we would need a scale that allowed us to estimate the generosity of each participant across a full spectrum of possibilities.
Why it matters if impartial altruism drives leftism
One reason you might think it matters if impartial altruism is a driving factor behind leftism is that it gives a meta-level argument for leftism, above the level of any individual policy. The people who actually want to help people are more likely to be the people who are right in any given debate about how we should help people.
Another reason you might think it matters is because it gives us insight into what drives political conflict. I’ve been banging on about this for years. There is a theory of political conflict according to which it is driven by disagreements about the facts. Obvious differences on factual questions come into play, but I do not think this is the primary driver. I think the primary drivers are 1. Differences in values 2. Differences in social location shaping self-interest and perceived ingroup. If the left and right vary greatly on an important value (altruism)- perhaps the most important value- this gives evidence against the ‘mistake’ view and for the ‘conflict’ view.
Finally, you might think it matters because you’re leftwing, and you want to know if there are any levers you can pull to make people more leftwing. This study doesn’t answer that, but it does suggest that if you could find a way to increase impartial altruism, maybe you’d thereby have found a way to make people more leftwing.
Income, giving, political views and other factors
There was a significant relationship between giving and income, but almost no relationship between political views and income, so little risk of a confounder.
There was a large relationship between education and giving. Whether this was mediated through income and education’s effects on political belief, or whether education has an independent effect on altruism needs to be determined.
The genders were essentially the same.
I couldn’t find a relationship between either degree of self-rated religiousness/spirituality or Christian beliefs, and the portion of money given to charity.
Ultimately, this study is purely correlational, if anyone wants to try and do causal inference using a properly controlled multiple regression, let me know and I’ll send you the data. Alternatively, if you just want to run a correlation, let me know and I’ll run it.
Trolley problem: Lever puller versus non-lever puller on generosity
From a philosophical perspective, it is not right to treat the trolley problem as a measure of consequentialism, or rather, it remains to be established that it measures consequentialism because, from a philosopher’s perspective, it does not have face validity. This is because while almost all consequentialists would pull the lever, many non-consequentialists would too- e.g. on the basis of an argument that the death of the one is not intended, merely foreseen. Exactly what the trolley problem measures is difficult to say, but perhaps I should have used the fat man problem instead.
Still, for what it is worth, there is a significant difference- both practically and statistically significant- between those who would pull the lever and those who wouldn’t. P = 0.0033, odds ratio 2 to 1, those who don't pull the lever are twice as likely to give less than 25 dollars.
My guess is that there’s a fundamental difference between those who think there is a moral distinction between killing and letting die and those who don’t, and this difference correlates with your answer to the trolley problem, and your donating habits. If you think letting die is nowhere near as bad as killing, all well and good- charity is somewhat optional. If you think letting die is as bad as killing, or nearly, then you could be very culpable indeed if you don’t give to charity- hence our results.
Bears
People who disliked or were neutral on bears were 50% more likely not to give at least 25 dollars than those who liked them. This effect was significant P = 0.0094, though nowhere near as strong as the political effects. This may have been in part because people who like bears are more leftwing?
Lie scale
In psychology, there are questions called ‘lie scales’. These questions involve asking the participant whether they have met some insanely high moral bar that very, very few if any people have met, on the assumption that if they say yes they are probably lying. My sense is that, in many cases, they are not assessing the tendency to lie per se, but the tendency to distort one’s own past in one’s mind.
Those who said they had never lost or broken anything belonging to anyone else were about twice as likely to be ungenerous, and this effect was significant. Those who said they had never said anything nasty about anyone else were twice as likely to be ungenerous, but this effect was non-significant because the sample size was so small.
If I am right that these scales are not really assessing lying, but self-deception, this raises an interesting question- would giving people clarity over their own past and their own actions increase generosity? I am not the first to suspect that the failure to confront oneself impairs altruistic love for others.
People who thought they’d never broken or lost anything belonging to anyone else were also about twice as likely to be rightwing as other respondents.
Further research on the relationship between lie scale questions, altruism and political beliefs would be welcome.
MDMA
A history of having taken MDMA, disappointingly, did not predict altruism. However, the survey was underpowered with respect to participants who had taken it ten times or more (only 28 respondents, and only 11 who had taken it more than 30 times) so never say never I guess. If anyone has any idea how I could get in contact with people who have taken MDMA at least 30 times, let me know!
Overall reflections
I chose a 3:1 ratio (for every dollar you forego, you can give three dollars to charity). to try and elicit equal amounts of altruism and selfishness. It turns out I underestimated the charity of the sample I think a better ratio for anyone replicating the study would be 2:1 since there was a strong imbalance towards people choosing to be generous, which I suppose is comforting, if not for theoretical reasons, than for practical. Making the ratio less lopsided by making the charitable choice less attractive. This should increase power. We had no problem with power whatsoever, but only because of our absurdly large sample size.
My initial plan was to use the impunity game, rather than charity, but for various reasons, I was worried about how people would interpret it in this context, e.g. “You’re asking me to give away some of my prize to a random stranger who didn’t win? Why?”. I think a follow-up study using the impunity game would be welcome. In part, this is because I am worried about comparability in giving. My initial hypothesis was that leftwing people are more inclined to impartial altruism than rightwing people and that this explains much of the left-right gap. However, charity isn’t a perfect measure of impartial altruism. What if leftwing people are giving to the local association for the protection of some godawful run-down park? What if right-wing people are giving to the patriotic-veterans association? I was tempted to specify a list of charities but was worried that some participants may have objections to each of them.
My final thought is that I am, as I so often am, struck by sublimity after investigating the dataset. Sociologists often talk of the crud effect- the tendency of everything to correlate with everything else. In truth, when I look at data like this, I am perpetually struck by the fact that most things, despite seeming like they should go to the heart of a person, don’t predict much variance. From people’s self-assessment of their morality to people’s self-assessment of their own gender- all these things are not especially predictive of things that you might naively think they should be. A few select things will be strongly related but features that seem like they should be deeply telling seem… mostly inert. There was one year of median age difference between those who gave it all and those who gave none. Even having children had no apparent effect on generosity, and one could think of all sorts of reasons it should come into play in both directions- think of the countless way children shape a life. I am not such a fool as to think this is a good argument, but sometimes I look at variables that should be related for a dozen dozen reasons, see little relationship, and find myself wondering if non-compatiablist free will exists after all.
I will investigate any hypotheses requested in the comments. The survey is now closed, and the prize winner will be selected on May 1st.
I think a bigger issue than theoretical altruism, is the issue of altruism at the point of a gun. I don't like using the bifurcated version of the political spectrum here but for the sake of the thing:
Are the left or the right more likely to raise taxes on individuals for whom that would be a bigger burden than on themselves?
Are the left or the right more likely to raise taxes on corporations because they think they are all rich and don't deserve the money they make and therefore 'the state' has the right to it more than those who earn it do?
Are the left or the right more likely to create an environment that supports, or at least doesn't overburden small businesses which create so much value for communities, large and small?
Are the left or the right more likely to support legislation that benefits out of state corporations at the expense of smaller local businesses?
Are the left or the right more likely to take away rights and then set up legislation that makes it almost impossible for them to be given back by a future version of the legislature?
Are the left or the right more likely to reform corrupt government institutions which cause so much harm and chew up so much local cash and provide a deeply degenerate version of the product they purport to produce or provide (thinking education in my town)?
Are the left or the right more likely to give substantive tax relief to families?
Are the left or the right more likely to leave the real altruistic giving to the individual rather than the hand that holds the power of life or death over your freedom and property?
I know the answer for my area, and these things are much more important, and impact the families and institutions in my neck of the woods MUCH more powerfully than the theoretical altruism of people willing to take the time to do a survey.
My biggest problem here is in accepting donation as a stand-in for impartial altruism, given the frequent -- almost to the point of cliche -- type of claim associated with the political right, that free money encourages counterproductive habits which wreck communities. A utilitarian who sincerely held that belief might very well personally sacrifice in order to *reduce* the scope of certain common human-to-human charities.