What is it about?
One of the most important functions of financial markets is to aggregate the traders' private information through prices. As the price of a security (e.g. stock, bond etc) goes up or down, it reveals the private information of those who buy or sell. In the long term, a well-functioning market aggregates all the available information, hence the price of the security (and its underlying assets) is the best predictor of its value. But there is a catch: we show that this long-term process only works for very few types of securities. This is not great for market designers, because it is not always easy to know which securities will work with which markets. How can we fix this?
Featured Image
Photo by Luke Chesser on Unsplash
Why is it important?
Our paper shows that, as the cost of information acquisition and processing goes down, almost ALL securities become better at aggregating information. This is huge for market efficiency! It is also relevant in today's world because the decrease in information acquisition and processing is accelerating with the advent of AI and the huge availability of data. The change isn't gradual - there is a sharp 'transition' where securities suddenly become good at aggregating information after information costs drop below a threshold. Think of it like this: Even a small decrease in information costs can suddenly make a market way more efficient at revealing the true value of assets.
Perspectives
Our paper makes several other contributions. First, some securities NEVER aggregate information well, even with super cheap information. Second, cheaper information does not always speed up price discovery. For some securities, it can actually slow things down! Our model is very general as we allow for strategic traders and prove results for all Nash equilibria. Despite combining both partitions & Blackwell experiments (which is rarely done in the literature), the paper delivers crisp results: a complete, easy-to-distinguish classification of securities into 3 classes. First, the ‘always separable’ securities aggregate information irrespective of who trades or what is their information structure. This class is very small and uninformative. Second, the securities which aggregate information if the cost is sufficiently low. This is the most generic and informative class of securities. For example, a security that pays differently across all states of nature. Finally, there is a small class of securities that may fail to aggregate information, even if the cost is very close to zero. Surprisingly, these three classes are easily distinguishable just by looking at the payoff structure of each security.
Spyros Galanis
Durham University
Read the Original
This page is a summary of: Information Aggregation with Costly Information Acquisition, July 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3670865.3673502.
You can read the full text:
Contributors
The following have contributed to this page
