Is Loss Aversion a Thing?

One thing I like about working in academia is the variety of styles of corridor talk you encounter that challenge your perspective on things you think are most likely to be true. I have tried to keep a fairly Humean view on things and not get too personally wedded to empirical generalisations. Having said that, I would still be very surprised if I get to the end of my life not believing that default effects matter in several circumstances beyond purely incentive effects, or that people make decisions on financial products based on confusion at least partly engendered deliberately by the provider or clumsy regulation in several cases, or that people will generally be more sensitive to time distance of rewards at near intervals compared to far intervals etc., And for the most part but possibly to a lesser extent (as it frankly is a more abstract idea than the others) I would be surprised if I didn't believe in some version of the idea that losses matter more than gains in terms of valuation and decisions. 

So it was good that my colleague Fred Basso sent me several papers questioning the basis of loss aversion, including this very thoughtful paper by Yechian (2018) arguing that the evidence for loss aversion that existed around the time prospect theory was published was unclear. A 2018 Journal of Consumer Psychology dialogue started with a paper by Gal and Rucker with the provocative title "The loss of loss aversion: will it loom larger than gains?"

A key idea in the Gal and Rucker paper is that many of the effects associated with loss aversion are only weakly associated with the core idea. And that when you think of things like the endowment effect as separate aspects of behaviour that themselves can be impacted by many things such as reference points with respect to market values, strategic behaviour, etc., then the case for loss aversion as a primary influence on human behaviour becomes a lot weaker. 

The core paper in the journal is followed by a number of responses, including the one below and here by Simonson and Kivetz largely arguing that while loss aversion has a number of potential moderators that should be developed further, the case for its demise is overdone. 

There are a number of recent papers that provide a more comprehensive summary of the empirical case for loss aversion than existed at the time of the above critiques. The Ruggeri et al multi-country study from 2020 replicates most of the original propositions of prospect theory though, as acknowledged by the authors, this does not itself demonstrate that loss aversion itself is the driving factor for the decisions seen in prospect theory set-ups. 

Most impressively is the recent meta-analysis of 607 studies of loss aversion by Brown, Imai, Vieider and Camerer. This work conducts a very systematic trawl through empirical estimates of loss aversion and includes controls for things like study quality and potential publication bias. Their work suggests a loss aversion parameter of approx 2/1 of the type that is discussed often in the literature is a solid rule of thumb to describe patterns from the last 40 years of studies. 

I am not volunteering to do it with current other time constraints but it would clearly be interesting to go through the 607 studies identified by Brown et al from the point of view of potential moderators of loss aversion. They do this to an extent in their already very impressive paper but mostly from an international comparison, sampling design, and statistical quality perspective. There are also some interesting discussions to be had about the limits of our claims about human behaviour in a world where global high-quality probability samples allowing for measurement of the type needed to separate out moderators of effects like loss aversion are extremely difficult to achieve in practice.