Streetwise Professor

September 3, 2009

Agency Problems

Filed under: Commodities,Derivatives,Economics,Exchanges,Financial crisis,Politics — The Professor @ 3:25 pm

The joint CFTC-SEC meeting about how the two agencies can play better together has spurred numerous suggestions that the agencies merge, or better yet, that the government create an uber-regulator responsible for all aspects of regulation of financial markets and financial firms.

I’ve written before that this is a very bad idea, and in light of the return of attention to this issue, thought it worthwhile to explain again why I think that’s so.  I’ll also lay out an alternative regulatory scheme that I think makes sense than either the existing system, or the uber-regulator alternative.

The argument against a regulator with myriad responsibilities for myriad firms rests in economic theory, notably agency theory.  Avinash Dixit has laid out the argument in a nice book that is somewhat accessible to a non-professional audience, so I’ll just hit the high points.

The idea here is that a regulator is an agent.  Who is the principal?  Well, Congress is the primary principal.  Thus, fundamental mechanism design issues that arise in the context of agency theory as typically applied to a firm (principal) trying to motivate and control its employees (agents) apply to the Congress-regulator relationship too.

Dixit notes that Congress designs regulators.  Moreover, Congress, both through the statutes it passes establishing and governing the regulators, and through the oversight process, provides regulators with incentives.  In particular, Congress can subject a regulator to very strong, “high powered” incentives, or very weak “low powered” incentives.  Agency theory tells us that the power of the incentives will depend on the nature of the organization/regulator.

Specifically, agency theory implies that low powered incentives are generally desirable when an agent has to perform multiple tasks.  The more tasks, typically, the weaker the incentives.  If you provide really high powered incentives, unless it is possible to measure everything the agent does with high and roughly equal precision, the agent will have an incentive to game the incentive system and do the things that provide very high rewards, and neglect the tasks that don’t.  As an example, let’s say you let cops keep a portion of the stolen goods or drugs that they seize.  You can bet that the cops would spend all their time looking for thieves and drug dealers, and have very little incentive to solve murders or intervene in domestic disputes: where’s the money in that?  Moreover, it’s hard to measure the effort or effectiveness or benefit of murder solving or separating Mr. and Mrs. Smith, so it’s hard to design an incentive to reward what you can’t measure.  So, when that 911 call comes from the Smith residence, don’t expect much of a response: the cops will be out looking for the dealers.

The resulting allocation of effort can be very perverse.  To give an incentive for the agents to do things that are hard to measure, you have to provide low powered incentives.  You can’t reward drug seizures, because that would induce the cops to spend all their time looking for drugs, and little time policing other kinds of (less remunerative to them) crimes.  Thus, low powered incentives are necessary to induce an optimal allocation of effort across tasks.

The more alternative things that the agent can do–the more tasks he performs–the more possibility for gaming high-powered incentive systems, and typically the weaker the incentives have to be.  Therefore, it is typically optimal to subject an agent that is  responsible for many tasks with lower powered incentives.

If the agent has many principals, the problem gets even worse.  Each principal has an incentive to get the agent to do what that principal wants, at the expense of the other principals.  So, it may be optimal to subject an agent working on a single task to weak incentives (fixed salary, no bonus, 9-5 hours) if he works for many principals.

Put these problems together, and you usually have to impose extremely weak incentives in order to reduce opportunistic gaming of the incentive system.  A multi-tasking agent working for many principals should typically be subject to very weak incentives.  And having very weak incentives, he is unlikely to work either hard or smart.  That may seem bad, but it’s less bad than having the agent work hard and smart to game a high-powered incentive system.

This means that a big, uber-regulator responsible for regulating everything in the financial space would be a disaster if Congress subjected it to strong incentives.  The uber-regulator would have many, many tasks to perform: oversight over security issuance, disclosure, bank capital and risk taking, derivatives trading practices, and on and on.  Moreover, it would have many principals.  It might seem there is only one–Congress–but Congress is not a unitary body.  It has numerous committees, many of which would have a piece of the responsibility for overseeing the agency.  Moreover, on a given committee, there are multiple Congressmen, each of whom is responsive to distinct interests unique to his district or state.

So, Congress would likely respond to this by providing the regulator with very weak incentives.

Thus, the alternative outcomes with an uber-regulator: an out of control body gaming high powered incentives, or a slothful, inefficient organization subject to weak incentives.  The likely outcome would be the latter, a living illustration of Eugene McCarthy’s statement that “the only thing that saves us from bureaucracy is its inefficiency.”

So what do I propose instead?  A more functional division of regulatory responsibilities, and one that mitigates the multi-tasking/multi-principal problems.  In a nutshell:

  • Split off the primary market, accounting, and disclosure parts of the SEC into a separate organization.  Call it the “Securities Commission” or the “Corporate Financial Information Commission.”  This agency would police fraud in the issuance of securities, and fraudulent disclosures and accounting practices.
  • Combine the market operation parts of the SEC and the CFTC into a Trading and Markets Commission.  This body would have responsibility for overseeing the operation of exchange and OTC trading mechanisms.  It would essentially be a “microstructure” regulator responsible for secondary market structure issues in securities and derivatives, manipulation and fraud in securities and derivatives trading, and the policing of agents who execute transactions for principals (e.g., stock brokers).  Under this system, there wouldn’t be the turf battles over the regulation of different kinds of financial products, or attempts to design products in order to have their trading overseen by one regulator or the other.  All trading would be under one regulator.  These trading-related matters are largely distinct from the information/disclosure related issues, and there is no good reason to combine all under one regulator.
  • Create a prudential regulator with the responsibility of regulating the risks of financial institutions.  This regulator would cover major financial intermediaries, regardless of corporate form, that potentially pose a systemic risk.  Thus, banks, investment banks, large insurance companies, and other levered institutions whose failure would pose a risk to the financial system would be subject to this prudential regulator

This division of authority along more functional lines would mitigate the multi-tasking and multi-principal problems that an uber-regulator would pose.  Of course, there would still be multiple principal and multiple task problems under this allocation of responsibilities, but the organizations would be more focused, and could be provided with stronger incentives to do their jobs than would be the case under a single, overarching regulator.

Is this likely to happen?  Surely not.  But hopefully people will start to awake to the fact that the seeming elegance and simplicity of one-stop regulatory shopping is actually a recipe for the next crisis.  Think that regulators failed last time?  Just wait until you have only one, and you’ll see a real failure.  Just like businesses are subject to diseconomies of scale and scope, regulators are too.  Sure, securities issuance and securities trading are related.  So are steel manufacture and automobile manufacture.  But it proved very inefficient to combine steel and car production in a single firm, and in my view it is very inefficient to combine all of the many aspects of financial market regulation in a single agency or commission.

Print Friendly, PDF & Email


  1. Very instructive post on principal agent problems. But the link to the Dixit book doesn’t seem to go to the book. I would be interested in knowing which book by Dixit you referred.

    Comment by Dan — September 4, 2009 @ 3:00 pm

  2. Dan–

    Thanks. Sorry about the link. Try this: The Making of Economic Policy

    The ProfessorComment by The Professor — September 4, 2009 @ 3:12 pm

  3. To throw in one more idea, make the regulators virtual wherever you can. Publish open standards (like TCP / IP protocol for instance) and use secure information exchange systems to verify if standards are adhered to. This would make regulation light and easy to modify and upgrade and compliance less expensive for companies. Auditors would lose their business and might not like it though.

    Comment by Surya — September 4, 2009 @ 4:22 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress