Froeb’s three questions for diagnosing how to improve decision-making in an organization are surprisingly hard to use. The second step (“information problem?”) tends to cause the most confusion. I’ve also added the implicit fourth step too.
- Who made the bad decision? Responsibility?
- Was there an information problem? Beware the hindsight fallacy explained below.
- Did the decision maker commit a fallacy and misuse or ignore useful information?
- Did the organizational structure keep useful information away from the decision maker?
- Did the decision maker have good incentives?
- Solutions: How do we change the three dimensions above to get better decisions?
Algorithm question #1: Responsibility
I am often struck by how hard this simple question is for many students to answer. People often want to be vague because of a desire to avoid blaming anyone, but if there is going to be any hope of change, the person or deliberative body that had responsibility=power for the decision must be identified.
This algorithm only works at an individual level, so if it is used to analyze a group decision like a board of directors, then it has to be repeated on enough members to change the outcome of the vote.
One of the reasons Americans dislike our national politicians is that it is nearly impossible to hold any of them accountable in our political system. For example, Trump’s signature policy proposal in 2016 was to build a big, beautiful wall on the Mexican border. Who gets the credit/blame for the amount of wall that was built? The House, Senate, President, state and local governments, or courts? They all share some part of the responsibility for how much wall building gets done on the border and it is really hard to know who to blame (or give credit to) for what happened. Nobody cared that Trump didn’t get Mexico to pay for it because nobody ever expects American politicians to follow through because politicians always just blame their failures on obstruction by opposing politicians and many campaign promises are more about theater than realpolitik.
In contrast, in a typical parliamentary democracy, there is nothing but the courts to hold back the ruling party and voters expect that politicians to follow through with what they say they will do. If British Prime Minister Boris Johnson doesn’t follow through on a campaign promise, he cannot blame anyone but himself because there are almost no checks on the ruling party’s power. Corporations are even more straightforward in holding employees accountable.
Being vague about who is responsible for making a bad decision can be good politics for both executives and politicians, but it makes it impossible to fix decision making problems.
Algorithm question #2: Information
The problem is how to know who has the wisdom to make wise decision makers and managers are by far the hardest people to manage. Fortunately, a true information problem is exceedingly rare because it can only be due to the misuse of available information. There are two types of information misuse:
- A fallacy: A abuse of available information.
- An organizational structure problem prevented the decision maker from getting available information that is needed.
A fallacy isn’t caused by a lack of accurate information, it is caused by faulty reasoning (or irrationality) that misuses the available information. Froeb is extremely misleading when he asks, “did the decision maker have enough information…?” because he is asking this question in hindsight, and in hindsight we always feel like we didn’t have enough information. If we had perfect information about the future, we could avoid all regret, but that is impossible.
Froeb’s framing of question #2 tends to mislead people because there is a natural human tendency to fall prey to the historian’s fallacy (a.k.a. “hindsight fallacy”) when thinking about past information problems. This happens because after a problem has happened and we know how it ends, we can’t think about the original context of the problem without the ending coloring everything about the original decision. It is like the way that watching a movie twice in a row is a totally different experience the second time. In assessing past decisions, it is very hard to avoid erroneously assuming the existence of information that wasn’t available at the time a bad decision was made.
For example, the victims who died in the World Trade Center on September 11 did not make a bad decision to go to work that day because they had no way of obtaining information about the terrorist attack before it happened. That problem was NOT due to an information problem. There can only be an information problem if reasonable people could have gotten better information to help make the decision. If there was no feasible information that could have led to a better decision, then it wasn’t a bad decision. You might as well say that gravity made a bad decision to collapse the twin towers on 9-11. Thinking that a bad outcome necessarily means that there was a bad decision is an example of the hindsight fallacy and that leads to bad management. Sometimes very good things happen because of very bad decisions and sometimes very bad things happen despite all the best decisions being made. We should ideally reward people for good decisions rather than lucky outcomes.
People make lots of mistakes because of lacking the correct information, but if they did not have the correct information when they made the bad decision, then they didn’t fall into a fallacy. As Froeb says, if there was no feasible solution, there was no problem! So if the decision makers didn’t have some information that later turns out to be crucial, then they can’t be blamed for not acting on information that was not available.
Let’s review the two main ways that true information problems occur.
1. People misuse their available information.
An example of this would be using payback period rather than NPV and ignoring the time-value of money. That means using a discount rate of zero. Another example would be falling into the sunk-cost fallacy.
2. An organizational structure problem prevented the decision maker from getting available information that is needed.
An example would be if someone in the sales division makes decisions about transportation without taking advantage of the knowledge that people in the logistics division have.
Algorithm question #3: Incentives
Most of the book focuses on selfish pecuniary incentives and its strength is in how to use them as well as they can be used. Monetary incentives can be dangerous because they often create unintended consequences. For example, in many of Froeb’s case studies, workers make bad decisions because they are too selfishly focused on monetary incentives. Organizations where workers have more intrinsic motivation to do good work could avoid those problems, and it is important to build an organizational culture where people generally want to take pride in what they contribute. Unfortunately, the book neglects crucial non-monetary incentives like inherent human drive to
- find meaning in what we do.
- be good people.
- have friends.
- seek beauty.
- become excellent at something that we do.
Within organizations, non-monetary incentives are more important than monetary incentives for guiding day-to-day activities. The culture of an organization is extremely important for shaping the norms of which incentives are emphasized and which are not. Froeb does not touch on organizational culture at all, but shaping organizational culture is one of the most important responsibilities of leadership. Leading by example and inspiration is one way to change a culture and punishing workers for being selfish at the expense of the organization is another way. An organization’s budget is another way of creating organizational culture and culture creates intrinsic incentives. A budget communicates priorities which demonstrate the organization’s values. Changing organizational norms and culture is a topic that comes up in leadership classes more than in economics, so this class will not focus on that, but as you read the book, you should critique Froeb’s peculiar view that humans are inherently selfish and obsessed with money above all other concerns.
Extrinsic incentives are difficult to design because of what computer scientists call the alignment problem. This is one of the dangers of artificial intelligence. Programmers can design artificial intelligence to reliably respond to the exact incentives that they design, but any incentives humans can design still produce surprising unintended consequences. For example, if we program robots to have a primary motivation to prevent humans from being harmed, then they might stop us from ever driving or crossing a road because of the risk of harm. Even when programmers can play God and design intelligent robots with custom-built motivations, that can still lead to disastrous outcomes!
Fortunately, humans are better at achieving alignment with each other because normal humans have an intrinsic desire for alignment with other humans. We experience it as a desire for feeling community, friendship, respect or love. Everyone realizes that money cannot buy love, so it should not be surprising that money cannot buy alignment in an organization either. Culture and values are what creates the best human alignment so that people work well together and an excessive focus on money can get in the way of our alignment with the people we need to work with and other intrinsic motivations.
Solutions: Changing the organizational structure or incentives to get better decisions.
The point of the algorithm is to change the organization so that decision-makers will perform better in the future. There are three common ways that correspond with the three questions:
- Can a different person be given the responsibility for making the decision (because someone else has better information and/or incentives)?
- Can the decision maker be taught a better algorithm for making decisions that will avoid their fallacious thinking?
- Can the incentives for the current decision-maker be improved? Don’t just consider pecuniary incentives and beware of unintended consequences. Every change in incentives has tradeoffs because it must shift resources away from something else the organization had been prioritizing. Examine the organizational culture and what incentives that are valorized and what are de-emphasized in the culture. Intrinsic incentives can be more powerful than extrinsic ones.
In my experience, most solutions involve incentives (question #3) and secondly organizational structure (changing the answer to question #1). Information problems (question #2) rarely yield solutions because it is hard to avoid fallacies as explained above. So the bad news about information problems is that they are rarely solvable. The good news about them is that they are very rare compared with incentive problems.
Froeb focuses on changing pecuniary incentives, but an organization’s culture produces powerful intrinsic motivation. Changing organizational culture can be difficult, but it is a better way to solve a decision-making problem if it can be accomplished. Organizational culture is determined by the norms and ethics of a group. If an organization is truly motivated by its mission, that eliminates the incentive problems that cause most of the problems in Froeb’s anecdotes. For example, in several of Froeb’s anecdotes, there are organizations that have problems because of managers and/or other employees who are lying and cheating their financial incentive system and reap selfish rewards. Froeb sees this kind of behavior as a natural part of human nature because he is a believer in the rat-actor paradigm, so he blames bad incentives rather than bad values, but when people are willing to lie and cheat, there are no pecuniary incentives that work well and it is probably best to fire the offender and send a clear signal that lying and cheating is not acceptable. Getting rid of egregious behavior helps reinforce a culture of honesty and teamwork for the remaining employees. A cheating coworker is toxic and most organizations immediately fire people who mislead others in order to cheat the organization.
Forgiveness can work if the cheater exhibits repentance and restoration, but that is beyond the scope of this class. Changing organizational culture is a key competency of leadership and this class hardly touches on it, but many leadership classes focus on these issues and I encourage you to bring your knowledge of leadership to bear in thinking about how to shape intrinsic incentives and organizational culture in solving decision making problems rather than just pecuniary incentives.