How Behavioral Biases Affect Executive Compensation
Humans can be pretty irrational when it comes to incentives. We favor smaller rewards now over bigger rewards later. Put a five-dollar bill in front of us, and we’d rather avoid its loss than gamble on its gain. And, as the Nobel Prize-winning economist Richard Thaler noted, we tend to accept or reject spending decisions based on how we’ve mentally earmarked the money—even though money itself is fungible.
These and other observations form the basis of behavioral economics. Classical economics assumes that people always act in their own best interests. Behavioral economics acknowledges that we, in fact, often don’t, and attempts to explains why.
So what does this mean for executive compensation? In a recent NASPP podcast, host Kathleen Cleary and I discussed how behavioral biases like loss aversion, mental accounting, and confirmation bias can make or break an equity compensation award. Put another way, the effectiveness of an award depends not only on real outcomes but also on how well the award was designed to take these biases into consideration.
In this article, I’ll walk through three examples of what Kathleen and I were talking about. Each example is set up as a problem for you to consider. Answer each question as honestly as you can, imagining how you might behave in real life. Then compare your answers to the common responses that experiments have shown.
At the end of each example, I’ll explain the behavioral biases at play and how they impact compensation.
Example 1: Accurately Estimating Ranges
Below are 10 questions. Since you surely do not know the answers (don’t cheat!), your goal is to state a range that you’re 90% sure encompasses the right answer.
- What is the word count of FAS 123(r) as issued? LOW: _____ HIGH: _____
- What was the income (using Summary Compensation Table conventions) of the median Fortune 500 CEO in 2012? LOW: _____ HIGH: _____
- What is the market cap of the hundredth-largest company in the S&P 500? LOW: _____ HIGH: _____
- How many people does Microsoft employ (as of June 2018)? LOW: _____ HIGH: _____
- What is the highest reported CEO pay ratio (CEO comp/median employee comp) through August 30, 2018? LOW: _____ HIGH: _____
- In what year did the Thirty Years’ War end? LOW: _____ HIGH: _____
- What is the orbital period of Saturn (in Earth years)? LOW: _____ HIGH: _____
- What is the distance from Phoenix, Arizona to Dubai, UAE in miles? LOW: _____ HIGH: _____
- What is the top speed (in mph) of an F-15 Eagle fighter jet? LOW: _____ HIGH: _____
- How many calories are in a McDonald’s meal consisting of a Bacon McDouble, medium fries, and a medium vanilla shake? LOW: _____ HIGH: _____
Compare your answers to the answer key (linked above, and also at the end of this article). If your range contains the answer, give yourself a point. The idea is that since you were 90% sure about your ranges, you should get exactly 9 out of 10 points. But don’t fret if you missed the boat! Most people facing a test like this get no more than five questions within the target range, and few ever get exactly nine.
Issue at Play, and What We Can Do About It
Overconfidence. The fact of the matter is that we humans are miscalibrated on probability, and we tend to be too confident in our predictive ability and knowledge. In plainer language: We’re not good at understanding the odds, and tend to overestimate the chances that we are right. This makes us very poor judges when setting up targets and thresholds.
One solution is recalibration. This relies on comparing more nebulous concepts to known, intuitive events—like a coin flip, picking a number from 1 to 10, or rolling a 6 on a die.
Alternatively, by introducing additional information, you might quickly be able to refine your estimate. For instance, if I told you that the distance from Los Angeles to London is 5,437 miles, that just might change the way you think about the distance from Phoenix to Dubai.
Long time horizons cause additional reason for concern. A particular place we routinely fail in prediction is looking beyond the early events. For example, experimenters have asked individuals how they think they will feel if their team loses the big game and how long the pain will last. People predict a long period of suffering and fail to realize that a nice walk the next day will take away a lot (albeit not all) of the sting. (As an avid Cubs fan, I can safely say that I’m almost fully over Game 6 of the 2003 NLCS after almost 15 years, even if the phrase “Bartman game” still gives me chills. )
Similarly, companies may be unable to look beyond their current strategy, such as how they might improve their product if a competitor steals market share, and the revenue gains that could yield.
In the executive compensation space, relative awards (such as relative shareholder return or performance vs. an index) provide a potential fix, because the goals set themselves relative to an objective benchmark. However, for revenue, EBITDA, and other targets, we recommend analyzing past data on actual outcomes for your company and others to assist with calibration. The more relevant data you can use, the better your results will be.
Example 2: Evaluation of Drug Side Effects
A drug is available that helps to prevent osteoporosis. Statistics show that of every 50,000 people who take it, 1,000 will prevent severe hip fractures that would otherwise occur. One will develop a severe fracture of a type which does not occur with any frequency without the drug. Would you take this medication?
With drugs of this type that are readily available, adoption rates are typically substantially lower than medical recommendations—many people will ignore doctor’s recommendations to take a drug they perceive as dangerous. These issues are typically seen around preventative care medications, procedures, and vaccines, and can even impact the behavior of medical professionals.
Issues at Play, and What We Can Do About It
Loss aversion. People are more heavily impacted by perceived losses than perceived gains.
Losses can be easy to disguise. Change the above example to say “49,999 of 50,000 patients experienced no negative side effects,” and see how different the same piece of information feels.
Know that employees feel more frustrated if they feel they missed out something they should have earned—or, worse, feel they were promised—than they will be pleased if they get an unexpected gain. Framing plans based on what you expect to pay out if things go “to plan,” rather than as an overall maximum, may reduce risk-taking. But it can also cause less friction if payouts fall short of the maximum.
Commission vs. omission. People are generally more upset when something happens because of something bad they did (e.g. taking a pill or selling a stock before it takes off), and less upset when the impact is a result of not taking an action (e.g. failing to act on a tip and buy that stock).
To avoid biases related to commission, it’s helpful to nudge employees in the right direction. A very common example is automatic enrollment and increases in 401(k) plans.
Overweighting and misunderstanding of small probabilities. People expect rare occurrences to occur far more frequently than dictated by probability.
Can you call 15 coin flips in a row correctly? A half-dozen die rolls? That’s more likely than being the one of 50,000 patients to experience the fracture. For an exact comparison, try guessing the last five digits of my phone number, with the only hint being that the last number is odd. If you guessed 81175, congratulations! But if you thought that was nearly impossible, there’s a lesson to be learned: When people have known comparators like this, they usually do a better job at conceptualizing probability.
In one of my favorite examples, drivers at the top of a hill with a bridge at the bottom are told that 1 out of 10,000 cars crash at the bridge, and the last 9,999 have passed safely. Of course, this last bit of information is useless, but everyone slowed down. One astute taxi driver took the information to the next level: He let someone pass him, and then sped down behind the other driver. I like to imagine his shock at the other driver’s survival.
Example 3: To Play or Not to Play
Consider the gambles below. Which bets would you take?
A. I’ll flip a coin. If it comes up heads, I’ll give you $200. If it comes up tails, you owe me $100.
B. The same wager as in A, but made twice in a row on separate coin flips.
C. I’ll spin a wheel: 25% of the time, you win $400; 50% of the time, you win $100; 25% of the time, you lose $200.
Most people faced with gamble A reject it, and those who reject it are quick to reject gamble B. Gamble C, however, is more appealing, and many more people are willing to take this than gamble B. With a bit of simple math, however, you can quickly see that B and C are the same wager. (If you were willing to take A without thinking, you may be a natural gambler—or, in psychology-speak, “less risk averse.” You can change it up by making the win in gamble A $125, and the wins in gamble C $250 and $25 respectively.)
Issues at Play, and What We Can Do About It
Narrow framing. When presented with choices or information, we tend to think of each piece in isolation. With that said, gamble B looks just like gamble A, just played twice. Only when all of the information is placed as a single event does our brain make the overall quality of the wager clear to us.
It’s important to make sure that no one looks at components of work or compensation in isolation. For example, a great bonus coupled with no raise might have different effects if they aren’t conveyed at the same time or on the same sheet of paper. Careful framing of job and compensation elements as parts of a whole set of responsibilities and decisions can help to improve job satisfaction and develop better career messaging to employees.
Of course, the above are just a few examples. Identifying how a program is truly working takes a thorough review of the program, its intentions, and prior results…and recalibration based on that review. We encourage everyone to take a thoughtful look at their compensation programs and see if they are indeed resulting in the right outcomes for both the company and the employees.
The NASPP podcast is called “#61: Irrationality, Behavioral Biases, and Award Design with Josh Schaeffer, from Equity Methods.” It’s 25 minutes long, and you can listen to it here.
Meanwhile, we welcome your feedback—and look forward to opportunities to discuss topics like this as they relate to issues you’re facing with your award program.
Answer Key to Example 1
- What is the word count of FAS 123(r) as issued? 115,071
- What was the income (using Summary Compensation Table conventions) of the median Fortune 500 CEO in 2012? $10.5 million
- What is the market cap of the hundredth-largest company in the S&P 500? $55.95 billion
- How many people does Microsoft employ (as of June 2018)? 131,000
- What is the highest reported CEO pay ratio (CEO comp/median employee comp) through August 30, 2018? 5,908 to 1
- In what year did the Thirty Years’ War end? 1648
- What is the orbital period of Saturn (in Earth years)? 29 years
- What is the distance from Phoenix, Arizona to Dubai, UAE in miles? 8,295
- What is the top speed (in mph) of an F-15 Eagle fighter jet? 1,875
- How many calories are in a McDonald’s meal consisting of a Bacon McDouble, medium fries, and a medium vanilla shake? 1,380