It’s widely known that financial system is very complex. Even just the credit system is complex:
A couple of years ago some CS theorists started analyzing one specific class of complex financial products—collateralized debt obligations. (The latest update of the paper is here.) Their primary focus is on the asymmetry of information between the buyer and the seller—the fact that the seller knows what among the debts being sold is bad and what’s good, whereas the buyer may not. What they prove is that the seller can manipulate the CDOs in such a way that it is computationally intractable for a buyer to know whether the CDOs they’re buying are junk or not. Worse than that, they show that such tampering can’t be detected after the fact, making them difficult or impossible to regulate; as they put it at the time:
Would a lemons law for derivatives (or an equivalent in terms of a standard clauses in CDO contracts) remove the problems identified in this paper? The paper suggests a surprising answer: in many models, even the problem of detecting the tampering ex post may be intractable.
While their results are fascinating, the complexity issues that they address are of the computational complexity variety rather than an examination of systemic complexity. That is, they looked at the downsides of the complexity of a specific financial product, not the downsides of the complexity of the whole financial system.
It’s this latter issue I’ve been trying to understand better. I’ve been on the lookout for ways to analyze and understand systemic complexity, but haven’t found much well-established work. (Not being an economist, it’s quite possible I’m just unaware of some well-known approaches here.) My hope was that something like the Ratnasamy complexity metric (developed to analyze the complexity of network protocols and distributed computer systems) could be applied to systems like the credit system above, and would enable economists and regulators to precisely study the complexity of those systems that succeed as well as those that fail. The metric yields an asymptotic complexity measure that is an attempt to capture how many pieces of information are manipulated by how many parties how many times. As it turns out, the complexity measure that this analysis yields happens to align with intuitive notions of complexity, and may naturally map to the analysis of systems like the credit system as depicted above. My hope is that by analyzing the complexity of a financial (sub)system, it would be possible to identify best practices—i.e. match those systems that work well with their complexity to see if / how the two are correlated—to reduce complexity and avoid future global economic meltdowns. That is, establish complexity limits beyond which regulatory bodies are allowed to step in and decrease system complexity.
As a final step, it would be valuable to develop simpler, more intuitive thresholds that are functionally equivalent to more complex thresholds that come out of the metric(s). Then we could both assure ourselves that the limiting mechanism in place would do the job and we’d be able to understand what the limit is. (We don’t want to rely upon arcane rules in an effort to make things more simple.) For example, suppose we were to institute two rules: “1. No financial institution may be responsible for more than 1% of national assets. 2. No financial institution may sell or transact more than two classes of products.” The first rule would help decrease the cascading damage caused by failures and the second rule would help increase the diversity and decrease widespread common-cause failures. Hopefully it’d be possible to relate these rules or ones like them to the system’s complexity analysis.
Stepping beyond the financial dimension, I’d be interested to learn if any such studies have been conducted on ecosystems. Many if not most ecosystems are complex, and involve an intricate dance of creatures and biogeochemical cycles, and so it seems somewhat unlikely that complexity itself is a core problem in unstable ecosystems. Are those problem factors discoverable from a macro analysis of the ecosystem in question? Is diversity more primary in this context than complexity? Do the patterns generalize?
At this point fixing the financial system seems like wishful thinking, but I still believe understanding the problems we’re in today in greater depth is the first step to figuring out how to prevent them from happening again.