In his book Predictably Irrational Dan Ariely described loss aversion by means of a set of experiments that go something like this:
A- take a person and promise her a substantial reward against completing a set of tasks. Measure stress level during tasks execution
B- take a person and give her the substantial reward up-front, but take back part of the reward for each task she fails. Measure stress level during tasks execution.
Ariely experiments showed stress was significantly higher with subjects that were in Experiment B. So much so that one of the subjects took the money and ran away by jumping out of a window. Ariely attributed the difference in stress levels to "Aversion to loss".
There is a difference between the scenario depicted by the experiments, and the cases in the OP's article. IN the article the author compares DMU as it changes from, say, 10K->2K vs 10K->18K, and argues that the latter has less impact than the former. Whereas in Ariely's experiments it is really comparing 2K-10K vs 10K-2K and showing that even if the Delta is the same, and both points are the same, subjects still experienced a different level of emotional distress due a visceral aversion to loss.
In summary, what is often described as "loss aversion" is actually just an expected property of diminishing marginal utility, and does not require a new term to explain it.
Loss aversion, instead, is state dependence that makes you feel worse about state X if you came from state X+1, and better about it if you came from state X-1 (assuming X is, say, wealth).
It seems to me that "loss aversion" might be reconstituted as "utility hysteresis" to avoid the ambiguity.
I would say that, at least academically, the difference between 'diminishing marginal utility' and 'loss aversion' is appreciated. To quote course material, 'When directly compared against each other, losses loom larger than gains'.
The difference in perception on the gain and loss side has been tested some, and the curve is supposed to look like this: https://goo.gl/PS8Z7c
"Loss aversion" should simply be seen as an observable behavior regardless of its cause. It makes no sense to say behavior stemming from diminishing marginal utility is not loss aversion if it leads to the same symptoms as "true" loss aversion.
His argument would be correct if the effect size were tiny, i.e. people got as much satisfaction from gaining $102 as from losing $100, but it's a much bigger effect than that. E.g. I will spend half an hour searching for a lost $5 widget (e.g. a lens cap or an iPhone recharging cable... ok a $20 widget) when I could easily earn far more money doing almost anything else (or keep my extremely expensive free time).
Also, recouping a loss is far more satisfying than an equivalent windfall, and DMU completely fails to explain that.
Ugh. For a person with $1,000,000 in assets, the difference between +/- $100 in marginal utility is almost zero.
Yet all the psychological studies will show that said person HATES losing $100 much more than they like gaining $100. It's a major effect.
Heck, if you even give people the same amount of money but FRAME it as a loss versus a gain, people change their behavior. The marginal utility is identical!
Loss aversion is MUCH larger than any difference in marginal utility. It's a real thing, that has huge effects on our politics, our economics, our media and our entire lives.
TL;DR: Article writer has absolutely no idea what he is talking about.
Agreed; the author is wrong. But his insight is that DMU is related to loss aversion when the sums involved are large, and that really is interesting. Hadn't occurred to me. He just overstates the case.
I think you're substantially misreading. The author is saying that people describe loss aversion in ways that are not obviously distinct from DMU. He does NOT say that they're the same. What he says is that most definitions do not distinguish loss aversion and DMU.
You give an example where DMU doesn't explain the phenomenon, so we can see loss aversion. He might be happy to accept your example. But that doesn't mean that in general you don't have to point out that they're distinct, because your case is rather extreme.
If you come by and swipe a quarter off my desk, I'm going to be way more pissed than if you came by my desk and left a quarter.
The delta of my being pissed is way bigger than the DMU of +/- $.25, or my feeling if I'd parted ways with that $.25 in a variety of other manners.
I don't know if what I'm experiencing is or isn't "loss aversion". But it's not "just DMU".
This is an important distinction. In my experience, people equate "loss aversion" with "unwilling departure of money" and "windfall" -- i.e. being unwilling to be swindled or ripped off.
I don't think your anger is a result of loss aversion, but rather the fact that someone committed predation upon you.
Civilization depends on an equilibrium where I don't steal your property, and you do not steal mine. Civilization flourishes when everyone is in the cooperate-cooperate quadrant of the prisoner's dilemma. When someone steals from you, they have violated this very basic principle. If you accept this defection without responding, then they will likely only defect more in the future. Thus maintaining the equilibrium requires you to respond harshly to even small defections.
We use thought experiments about magically appearing money, but the aim would be to later try to apply those ideas to more realistic cases of, for instance, risk-taking.
Interesting. A way of testing the author's assertion would be to ask: who feels worse, a person who gains then loses $1000, or a person who loses then gains $1000?
If there is a difference, then the order of events has impact beyond the initial/end states.
I'd also add that loss aversion's supposed "reference dependence" is addressed by a more fundamental concept, a cognitive bias: anchoring[1].
It's interesting how these (especially pertaining behavioral economics and behavioral studies in general) create compound ideas from combinations of various fundamental ideas. Sometimes, it adds value. Other times, it simply obfuscates the truth.
It makes sense to me that particularly 'salient' things get their own dedicated category, despite being a subset of some other category.
Consider racism. Most (or at least much) of the time it's really just a more salient and specific version of the 'in-group bias'. And I'd argue that on an academic level, researching and discussing it as the latter is vastly preferable to using the much more loaded concept of racism. But on a societal level it perhaps makes sense that we treat 'racism' as a category in itself.
Personally I think discussing whether x is 'just' a subset of y is less valuable than carefully delineating the contexts in which we use particular terms. For example, I wish that on a societal level we'd consider 'innate' differences a bit of a taboo, but that on an academic level we could go wild researching, say, IQ differences of particular populations without it being coopted by politically motivated individuals.
I vaguely recall Steven Pinker (almost?) making that argument in The Blank Slate, and even though I've never read Anathem, I understand one theme is the idea of having academics locked up and doing research similar to orders of monks isolating themselves from the world.
At the risk of this becoming (even more of) a ramble, I can't help but wonder what effect our increasing interconnectedness and the inability to do something in isolation has on all of this.
It's not as much a matter of determining whether "x is 'just' a subset of y", as it is a matter of finding the most fundamental truth about something.
For instance, if you state that (paraphrasing) racism is wrapped up in 'in-group bias' and it turns out that people are actually racist because of something in our DNA (similar to being afraid of spiders) or something like that (NB: completely contrived, to provide context related to your comment), the you would have lost the truth in a higher abstraction layer, by hiding the underlying principle.
Now, abstraction is a helpful tool, but it should be used to abstract things to the level necessary to convey an exact message. Einstein might have said, "Abstract things as high and low as necessary, while preserving the truth, but no higher or lower."... but in the meantime, I made that quote up.
Take, for instance, my "Anchoring" example. There are more fundamental things that are going on beneath the "Anchoring" abstraction. Perhaps a specific part of the human psyche, which is also responsible for recognizing patterns (e.g., Reticular Activating System) is also responsible for the mental constructs that lead to such a cognitive bias. But, once that science is understood sufficiently enough to be trusted, a concept like "Anchoring" is a helpful delineation.
In the aforementioned case of racism that I contrived, the abstraction layer of "racism" and the perceived lower level abstraction "in-group bias" hides the truth, and this fallacious abstraction actually becomes, as Noam Chomsky might suggest, a part of our mental grammar, which prevents us from ever learning the truth, unless the abstraction is broken in our minds.
I agree with you within the 'academic' context, but that doesn't address my main issue of whether this same approach can or should be used in a 'broader' context.
Could it be that on societal level creating a separate category like 'racist' is important and valuable, even if it's technically a 'fallacious' abstraction? We don't always have the time to properly get to the bottom of things, assuming all of us are even capable of doing so, so we're going to grasp for salient categories anyways.
It vaguely reminds me of the discussions here about functional programming versus object-oriented programming. Alongside the heated debate for and against either approach (and the definitions of the 'true' version of each), there's always someone who points out something like 'closures are just a poor man's objects', and 'objects are just a poor man's closures'.
Even if, for the sake of argument, the differences between FP and OP are not as fundamental or clear-cut as they seem, I'd argue that for me and many others these discussions are very useful. I'd never get out of my 'OOP box' to explore FP if it wasn't contrasted and put in a whole separate category and given a bunch of tantalizing pros that pull me to investigate.
Isn't 'truth' really just another abstraction, but one we cannot (yet) dive into to find the underlying 'truth'? What I mean is, we stop at a certain point not only because it's where we end up at for the time being, but also because it's a useful abstraction.
And to be clear, I'm arguing this primarily in the context of the already very murky and messy field of the social sciences / psychology, where definitions are a lot less definitive than in physics or mathematics, and where they have a much more immediate effect on society.
(honestly, I'm not sure I'm disagreeing with you, and I'm sorry if I'm perhaps not making much sense. I'm not usually this openly... explorative in my comments here. Your comment(s) just tickled my brain in a good way.)
Thanks for the reply. I just now saw this (haven't been nearly as active on here for the past week).
I enjoyed reading your reply. I had just one comment back:
> Isn't 'truth' really just another abstraction, but one we cannot (yet) dive into to find the underlying 'truth'?
It's not, because 'truth' is a logical construct, an a priori kind of concept, so whatever that 'truth' is that we find underneath what was once thought to be truth would just be a more fundamental truth (e.g., we discover that photons are actually an imbalance in another dimension or something).
I think the article is confusing. Loss aversion is a property of human decision making, while diminishing marginal utility is a property of some economic system.
The two may be the same, if you believe in (or talk about) subjective utility. But I think subjective utility is a terrible concept to begin with, so better not to go that route.
It seems to me that you can reverse the sense of the author's conclusion by changing the initial conditions:
The change in utility by going from 2k up to 10k (~ +1.3) is significantly more than the change in utility by going from 18k down to 10k (~ -0.55).
In the simple model provided the change in utility caused by a change in wealth, positive or negative, depends on your starting wealth. This is different to the "pop definition" of loss aversion, which seems to be making a claim that the ratio of the change in utility between gains and losses is approximately independent of your starting point.
I should point out that I don't know what the correct formulation of loss aversion is, it just seems to me that the argument presented is a bit weak.
A- take a person and promise her a substantial reward against completing a set of tasks. Measure stress level during tasks execution
B- take a person and give her the substantial reward up-front, but take back part of the reward for each task she fails. Measure stress level during tasks execution.
Ariely experiments showed stress was significantly higher with subjects that were in Experiment B. So much so that one of the subjects took the money and ran away by jumping out of a window. Ariely attributed the difference in stress levels to "Aversion to loss".
There is a difference between the scenario depicted by the experiments, and the cases in the OP's article. IN the article the author compares DMU as it changes from, say, 10K->2K vs 10K->18K, and argues that the latter has less impact than the former. Whereas in Ariely's experiments it is really comparing 2K-10K vs 10K-2K and showing that even if the Delta is the same, and both points are the same, subjects still experienced a different level of emotional distress due a visceral aversion to loss.