Laboratory versus Economic: Information Cascades
https://onlinelibrary.wiley.com/doi/full/10.1111/j.1540-6261.2007.01204.x
The theme of networks seems to be connecting seemingly abstract concepts of supposedly idealized scenarios yet still finding a practical way to apply them to real-world problems. While more general concepts like Bayes’ Theorem are more easily applied, other ideas like hubs and authorities seem intuitive at first yet doubtful to work smoothly when applied in practice, and yet the transition from paper to real-world is surprisingly smooth.
One such idea is the application of information cascades, a concept that essentially revolves around the idea that small samples of public data can change the choices of others, often overriding the more “logical” decision. Most of the time, the ideas behind information cascades are often tested in laboratory settings rather than in real-world examples. The most commonly done test is the classic urn problem, which is a variation of the “majority blue” problems that we are all aware of. Bayesian concepts allow us to figure out the logical probabilities, but after a select amount of people, these logical probabilities are eventually overrun, in fact, so much so that even if there is a change in what we might expect from an information cascade, they self-correct back to our original assumptions (https://academic.oup.com/restud/article/74/3/733/1563683).
With all of this in mind, the most common real-world example that many apply information cascades to are financial and economic settings. Stock or market events often mimic situations that resemble information cascades, where the information provided by one buyer will often influence the actions of others who are also interested in buying. While this seems like the most practical application of information cascades, they don’t actually act one-to-one. This paper (https://onlinelibrary.wiley.com/doi/full/10.1111/j.1540-6261.2007.01204.x) demonstrates that there is a discernable difference between how information cascades work in financial and economic settings versus laboratory settings. This may seem a little obvious in hindsight, but the final conclusion isn’t readily apparent. The main difference boils down to the fact that professionals are better at qualifying public signals, whereas student control groups are less able to conclude whether public information can be trusted or not. The immediate conclusion we can draw is that information cascades can’t be readily trusted, and while this is true, we can’t disregard the other interpretations of this paper. To fully understand this analysis, we would have to be able to qualify or even quantify the signals ourselves and then compare them to their private information. Being able to do that would reveal a deeper understanding of how many variables can go into predicting how the market flows in regards to information cascades. As of now, my current knowledge only knows how to analyze these values at a very basic or binary sense, but I imagine real-world examples are not as elegant.