Share This Page
Jordan Tong

Gaumnitz Award Winner Jordan Tong Shares Research Works in Progress

by Jordan Tong Monday, July 3, 2017
Assistant professor Jordan Tong poses for a portrait in WSB atrium

WSB Assistant Professor Jordan Tong won this year's Gaumnitz Junior Faculty Research Award. PHOTO: PAUL L. NEWBY II

Jordan Tong, professor of operations and information management at the Wisconsin School of Business, is the 2017 recipient of the Gaumnitz Junior Faculty Research Award. The Erwin A. Gaumnitz Distinguished Research Award series recognizes faculty excellence at the Wisconsin School of Business. Tong completed his Ph.D. in 2012 from Duke University. His work focuses on areas related to judgment, decision-making, and supply chain management. He spoke with WSB recently about some of the themes of his scholarship and aspects of current works in progress, including whether your boss may be underestimating you at work and why getting a second opinion might not be a bad idea.

WSB: Tell us about what you’re working on right now.

Tong: One of my current lines of work is focused on understanding how managers make judgements and decisions and how to improve them, especially in supply chain management contexts. I am taking an approach where instead of looking at the individual—whether the person is optimistic or pessimistic, or whether he or she likes taking gambles or not—I’m looking at specific problem structures that create tricky data that cause managers to be predictably biased. 

Let me give you a couple of examples. Managers often look at historical sales data to try to guess what demand will be in the future. The structure of the problem here can be tricky: sales data doesn’t capture the units you could have sold when you didn’t have enough inventory. Maybe some weeks you stocked out so the data isn’t reflecting that additional demand you could have had. So, unless managers recognize and correct for the trickiness of this data, they will underestimate historical demand which can bias their future decisions.

Another really tricky example occurs when managers try to pick the best product to develop based on imperfect estimates. Imagine you make estimates of how profitable five different products will be, and then you choose the one with the highest estimated profitability. Even if your estimates of the five different products are correct on average, the estimate of the product you chose will tend to be too high. This happens because accidently estimating too high on a product makes you more likely to choose it, but accidentally estimating too low doesn’t. So again, if managers don’t recognize and correct for this tricky problem, they will be overoptimistic in the products they choose and may subsequently overinvest. 

So, my general hypothesis is that any time the problem structure creates tricky data, managers are usually unable to fully correct for it. I think the flavor of my research lately is about identifying these really tricky and difficult problems out there in practice and trying to deeply understand their structure. Then we can better pinpoint when and where we should expect managers to need the most help and figure out how to help them.

WSB: Can you talk about some of the tests you are running currently? 

Tong: Sure, an easy one to explain would be a simple short experiment we did with jars and pennies that corresponds to the second example. We took five jars with a random number of pennies in each jar. Then we asked people to estimate how much they thought was in each jar. Of course, it’s impossible to guess exactly, but on average, giving them one jar at a time, they are remarkably accurate.

Now, put all five jars in front of someone at once. Ask him or her to pick the one with the most pennies in it and to guess how many pennies are in the jar. People overestimated how many pennies were in the jar they chose by nearly double.

Structurally, this problem is just like trying to estimate the profitability of five products, and then picking the one you think is best. And, as you might expect, we see similar patterns of results when we study those kinds of managerial contexts.

WSB: Why do you think this was happening?

Tong: It’s not because people are bad at guessing quantity, because they were fairly unbiased when given jars one at a time. It has to do with the statistical structure of choosing the one that you think is the best. You’ve probably erred high on the one you selected as best. So, unless you know about this statistical process going on, you will be left with this overly optimistic bias.

One thing I’ve experimented with is redesigning the system to avoid this problem. Mathematically, you can show that you should be able to reduce the overoptimism by using a second person. The first person chooses a jar he or she thinks is best. A second person is asked to estimate how much is in the jar. Why? Because you don’t want the person who chose the jar to also be making the estimate. We have found in experiments that this is in fact one way to reduce overoptimism. 

The design implications here suggest that you should get a second opinion from a person who did not choose—that second person is going to be less biased than the first person. Bringing it back to a sales and operations perspective, this suggests that you want to be careful whenever the person or team who chose which product to bring to market is the same person who decides how much the company should invest in that product. It’s probably wise to separate these two types of decisions.

WSB: Would you say that these ideas are applicable to other fields outside of supply chain?

Tong: Yes, they can certainly can be. Most of my research is motivated by problems in supply chain management, so I think that’s naturally where my interest originated. But once you start researching problem structures, my coauthors and I realize that the same problem structure can occur in many other places.

For example, as we mentioned earlier, let’s say your sales data doesn’t capture the demand that you could have sold had you had enough inventory. That’s what we call a censorship problem and it can show up in other contexts. For instance, my coauthor is currently working on a different context. If your boss assigns you a lot of work, he or she can always see how much you can complete. If your supervisor is not assigning you as much, you may just complete what you are given and not ask for more. Instead, you just go check Facebook or do an extra good job with what you’re given.

But as the boss, you are receiving the same kind of censored feedback as the sales data situation, right? On the days that the employee completed everything, you don’t know how much more he or she could have done. So, our hypothesis is that managers are underestimating the ability of workers who often complete their work. Conversely, if you’re that worker, you should be telling your boss, “I could have done a lot more!”

Read the papers: 

“Biased Judgement in Censored Environments” published by Management Science.

“A Behavioral Model of Forecasting: Naive Statistics on Mental Samples” published by Management Science.

This conversation also references Tong’s working paper “Good Choice, Bad Judgment: How Choice under Uncertainty Generates Overoptimism.” 

Jordan Tong is an assistant professor in the Department of Operations and Information Management at the Wisconsin School of Business.