Wednesday, April 14, 2021

The Local Maximum Problem

Ever since being introduced to this concept, I’ve been intrigued by it and see examples of it more and more throughout life, business, and public policy. This is the problem that occurs when people get stuck in a situation that is the best near-term or near-possible outcome but is not the best possible yet reasonable long-term outcome. 

The analogy is to imagine four people playing a game that has them blindfolded and linked arm-in-arm in a square configuration. Each member of this team is responsible for one of the cardinal directions (north, south, east, and west). Their goal is to locate the highest point possible. They experiment by taking steps to see if a step in that direction is up or down. If the step is down, they don’t take it. If the step is up, they take it. They keep walking until none of the four can make a step that is in the upward direction. This point is the conclusion of their game by reaching the local maximum. However it is most likely not the highest point on the surface where they’re walking. They just can’t reach (or detect) a higher point by virtue of their own rules. 

I believe governments are particularly susceptible to this problem. The rewards for experimentation that drive one out of a local maximum are very dispersed or completely irrelevant to those bearing the costs of experimentation. This is more than just people not wanting their cheese moved or having their apple cart disrupted. This is the very legitimate concern that an ambitious idea is going to have significant negative outcomes or the potential rewards will not accrue to those bearing the risk. It is an acute combination of asymmetric risk-reward and principal-agent problems.

The many, many public and private failures in the COVID pandemic are vivid examples. Perhaps the most costly in the United States were the CDC and FDA's insistence on using their own developed testing (staying with the controllable and familiar) and as important if not more so the refusal to allow challenge trials to speed the vaccine development process. Sadly this list goes on and on from "pausing" the Johnson & Johnson vaccine to not approving AstraZeneca's. 

The position those in power have taken are understandable but completely inexcusable. And we have ourselves to blame as these mistakes are just the latest examples of how the FDA works against medical advancement and is a deep net cost to society. 

To be sure individuals, firms, and other organizations are also susceptible to the LMP. Notice, though, the degree to which these entities are somewhat or greatly better structured and incentivized to resist and correct it.  

As a general rule, the more insulated and protected an entity is from competition, the more vulnerable they are to a local maximum. Hence, traditional banks are more vulnerable than are start-up fintech firms. 

To whom a firm or organization is held responsive has strong implications for its fragility to local maximums. As a firm is more responsive to those who reap rewards proportional to risk taken, it will better prevent the LMP. Hence, non-profits (highly responsive to donors rather than customers) are more at risk than are profit-seeking firms (highly responsive to owners and customers). 

Within a firm the dominant force becomes existing and entrenched stakeholders who are in comfortable, conventional positions. Hence, no one in marketing will ever suggest the firm experiment by not running ads

The degree to which a person faces public scrutiny or cannot capitalize on public adoration, the more they will rest once finding the local maximum. Hence, a public figure with a lot to lose/little to gain will tend to play it safe. 

Risk bearing requires compensation in the form of return, and this risk-return should be commensurate, symmetrical, and willfully accepted. Those are tough hurdles to achieve. All the more so when we are relying on force rather than persuasion. 


P.S. I believe Arnold Kling deserves credit for introducing me to this concept.