Tag Archives: behavioral economics

I made a custom major at Wharton: Behavioral Analytics

Can’t figure why this chart is bad science? You should probably major in Behavioral Analytics. (source)

At Wharton, I found myself taking lots of classes on behavioral economics and on data wrangling. There isn’t really a department that does both of these things, so I found myself somewhat at a loss for a major that encompassed these interests. Fortunately, Wharton lets MBA students declare their own majors – in fact, one out of twenty students do! So that’s what I did.

As my new major came up in conversation, a number of other students expressed interest in data and behavioral science topics. So many, in fact, that I led an initiative to get Wharton to officially recognize this major, and wrote an article about why we should do it (“Data is Sexy – let’s study it at Wharton”). Over 200 students signed a petition in support.

Hiring managers are increasingly looking for good data nerds, which is great for those of us who love this stuff. Hal Varian, Chief Economist at Google, has also said, “The sexy job in the next ten years will be statisticians.” So, not only are students interested in this, but companies are, too. Getting the major approved as an official option is still in process, but when I left a few months ago, it looked pretty promising for the classes of 2016 and 2017.

I’m sharing this here as an overview, because a lot of prospective and current Wharton MBAs have asked me about it, and this post will serve as a starting point / primer, as well as a description of what I did. The best part about declaring a custom major is you can, well, customize it! And you should. Prospectives who are reading this, think creatively about how to make your area of study really your own.

This post has a couple of parts:

  • Description of the major
  • The initiative to get it on the docket at Wharton
  • Some classes I recommend to Wharton students
  • Some ways for students to get involved in data at Wharton

Description of the Major Continue reading

When graded on participation, students tend to participate three times in class

One of my last activities here at Wharton was TAing for an undergraduate class.* Throughout the course, the professor asked students to tweet content relevant to the topics we discuss in class. As the TA, my job is to count how many times each student tweets about content related to the course, which will contribute to their participation points.

After learning about some cool distributions in Peter Fader’s Probability Modeling class, I noticed that the distribution of student tweets looked like one of the models we discussed – the Negative Binomial Distribution (NBD), which is a pretty neat distribution because of how flexible it is.

I counted the number of students who tweeted no times, just once, twice, etc. Here’s the original data:

Number of Tweets Number of students that tweeted this many times
0 134
1 14
2 6
3 11
4+ 6

I applied an NBD to this distribution. It turned out to fit pretty well visually, but the p-value was only 0.0005, which isn’t super great.

Screen Shot 2015-05-14 at 10.56.10 AM

However, the surprising part: in looking at the histogram, we can see that a disproportionate number of students tweeted about course content exactly 3 times; the model doesn’t fit very well there. We can account for that by adding in a spike at 3, indicating that there are a segment of students who are doing the activity exactly three times. Doing that, I obtained a ludicrously high p-value of 0.9712 – the model fits really well!

Screen Shot 2015-05-14 at 10.52.25 AM

One possible story, which I like, is that students believe 3 tweets is the optimal number to ensure that they get participation points, but tweeting any more than that would seem like they’re *too* engaged in the course.

Other math: I tried an NBD with just five buckets and a spike at 3 first, to not violate the E(x)>5 rule. The resulting (very high) p-value left me suspicious, so I wanted to try one more thing  – to account for the low degrees of freedom, I added in more buckets, despite E(x)!>5 . With 9 buckets, I still get a pretty sweet p-value of 0.8486. Here’s the data, and the histogram for that model.

Number of Tweets Number of Students that tweeted this many times
0 134
1 14
2 6
3 11
4 2
5 0
6 1
7 1
8+ 1

Screen Shot 2015-05-14 at 10.53.44 AM

Pretty fun stuff! Thanks, Wharton. :o)

*All identifying data removed! =)

The Default Option: “Plus” Gas? How gas stations might make additional $20k/year


The options at this gas station

The options at this gas station

The other day, my boyfriend and I stopped at a gas station just off the freeway in New England. A few moments later, he exclaimed, frustrated, that he couldn’t choose “Regular” gas – the pump automatically defaulted to “Plus.” He had to cancel out his transaction and start again to select Regular.

In psychology, studies show that setting a default impacts how people choose amongst options; they often tend to choose the default at a statistically significantly higher rate than they would have chosen otherwise. This is called the Default Effect.

Let’s take a look at how changing the default from “Regular” to “Plus” could impact the bottom line of this Shell station. Regular gas was $3.94 on this day, and Plus gas was $4.06. Let’s assume that this difference of $0.12 is consistent every day of the year.

Let’s assume that about 200 people come to a gas station on average per day. Of the 200, let’s assume that 75% of them would normally buy Regular, and are affected by this default change. Of the 150 remaining, let’s say 2/3rds of them cancel their transaction and change the default option from Plus back to Regular. So, 50 people per day will be affected by this changed default.

Assuming that the average transaction is 10 gallons per transaction, the gas station could gain $1.20 additional dollars per transaction by changing the default. (This is from 10 gallons x $0.12, from above).

An additional $1.20 per transaction for each of 50 people per day is an additional $60 per day. If a gas station is open 365 days a year (I’m assuming they’re open on holidays),  changing the default from “Regular” to “Plus” could generate an extra $21,900 per year in sales revenue for an average gas station.

Not bad for just changing the default option!


Behavioral Economics, according to Dilbert

Psychology Today has collected some pretty great Dilbert comics that explain some behavioral economics concepts. Here’s the full post, and a few of my favorites below.

Confirmation Bias. “Confirmation bias is the tendency to seek evidence consistent with a prior belief. In the above strip, Dilbert’s boss demonstrates this bias to a tee when he assumes his astute managerial skills are what caused a minuscule (and clearly unrelated) improvement in the company’s stock price.”

Losses Loom Larger than Gains. “In prospect theory (which is a key concept in behavioral economics), the pain associated with a possible loss is much greater than the pleasure associated with a gain of the same magnitude. In the above strip, Dilbert’s garbage man clearly understands this concept much better than Dilbert.”

Why No One Wants To Say “I Love You” First

Candice Chung wrote a great article in DailyLife titled “Why Smart Women Go for Jerks.” In it, she creates a game matrix for saying “I Love you” first in a relationship. Here’s an excerpt from the article, with the game matrix.

Game theory: Thanks to Russel Crowe, most of us are now well-versed in how game theory works, if only in a pick-up situation.  At its simplest form (i.e. Prisoner’s Dilemma), the theory helps you decide on the “best course of action following your opponent’s choice” – with one important caveat: it doesn’t necessarily yield the ‘best’ possible outcome, just the most logical. (Say, what now?)


So for example, in the case of whether or not to say ‘I love you’ in a new relationship, according to Nobel-prize winning economist John Nash, the dominant strategy (ie. one that leads to minimal loss no matter what the other person does) is to stay silent (lest one person confesses their love and gets rejected). Even though the “best possible outcome” is actually for both parties to take the plunge and say how they feel.

Given this information, would you say “I Love You” first, or would you wait?

Game Theory Drinking Games

Tomorrow, I’ll be taking a final exam in Game Theory. In classic business school fashion, a post-exam celebration would hardly be a celebration without copious inebriation. But, this is business school – so it’s got a bit of a twist. I’ll keep you posted on the outcomes.

Introducing … Game Theory Drinking Games.

I’ve summarized the key points below, with names removed to protect the innocent. I can’t take credit for these amazing creations – these come from my esteemed fellow students.

Cluster President Likes to Drink.”

Basic Rules

  • Prior to the game, we determine our Cluster President (let’s call her Sarah)’s favorite number of shots is. But y’all WILL NOT KNOW THIS NUMBER.
  • We will hand each of you a slip of paper. You will independently write down your bid for how much money you’d be willing to pay to buy Sarah this specific, and unknown to you, amount of shots.
  • The lowest bid “wins” the right to buy Sarah that number of shots along.
  • As an additional incentive, the winner will get 3 shots paid for, courtesy of the cluster.

An Example Continue reading

Monkey Hear, Monkey Do – why the Bolshoi Theatre hides “Clappers” in the audience

If you’ve ever been to a concert, you’ve probably experienced an encore – that bit at the end where the audience can’t seem to stop clapping until the performer comes back on for another short while.

At the Russian Bolshoi Theater, this isn’t an accident. Dancers will employ the services of so-called “claquers” (from the french “to clap”). Claquers are audience members strategically placed to applaud enthusiastically for these performers, at the end of and during the performance. Other non-claquer audience members, upon hearing the applause, will join in.

Basically, these claquers are the ballet version of a sitcom laugh track.

Why does this work?

Applause spreads through crowds like a contagion, rather than a reaction to the quality of a performance or a performer. There’s not a strong relation between how good a performance is and how much the audience claps. It’s more a question of what everyone else in the audience is doing.

In short: people clap if other people clap.

As a business person, you can use phenomenon this to your advantage. Next time you’re going to propose an idea to a group that you’re not sure will buy into it, talk to a couple of them one-on-one before the meeting. Let them know what you’re doing, and ask if you can have their support. That way, when the time comes to present, you know you’ve got a few people on your side.