Suppose you’re playing a game where the goal is to accumulate as many points as possible. Now suppose your decisions — and only your decisions — control not only how many points you get, but also how many points the other player gets. Suppose further that at the end of the game, you’ll be able to cash in your points at a rate of 10 cents per point.
Now consider this scenario, one of a series of similar scenarios throughout the game: If you take 7 points, then the other player gets 9 points. If you take 5 points, then the other player gets 3 points. How many points would you take? Let’s make this a poll:
Now think about a different scenario, which I’ll present below.
This time, it’s not a game, it’s a one-time payout, but again you control the payouts. You choose between two options: You can get $500 and some other anonymous person will also get $500, or you get $600 but the anonymous person gets $800. Make your choice in the poll below:
This scenario isn’t much different from the first one, except that the payouts are bigger and it’s not described as a “game.” In both cases, the only way to maximize the amount of money you get is to ensure that someone else gets even more money than you. But if our poll responses match the results of a recent study, we should find a different result for each.
A 1992 study led by Max Bazerman asked the $500/$600 question to business students and accounting firm managers, and most respondents took the hypothetical $600, even though it meant they wouldn’t be getting as much as their imagined rival. Arguably this is the optimal response, since it maximizes the amount of money the respondent gets. Who cares if someone else gets more?
But if you think about this scenario as a game, more like our first poll, then the only way to “win” is to consistently take less money. If you’re asked to make a series of decisions like the 7/9 point — 5/3 point choice, if you choose 5 every time, at the end of the game you won’t have as much money as you would have made by choosing 7 every time, but at least you’ll have more than that other guy. But if the game has several rounds, there’s another option: you could alternate between choosing 5 and 7 points, and at the end of the game, you’d have the same amount of money as your opponent, so you wouldn’t “lose,” but you’d still make more money than choosing 5 every time.
Arthur Kennelly and Edmund Fantino presented this challenge to 200 college students with a couple extra twists: some students were told they were playing with a live human in an adjoining room, while some were playing against a computer. Some students were rewarded with real money — 7 cents per point — while others were given the same course credit for participation, no matter how many points they earned. Here are the results:
As you can see, respondents fell into three distinct categories. There was a large group of students who always took more points for themselves and another group who always took fewer points — and therefore more points than the other player. But there was also another group that chose more for themselves exactly half the time, thus matching the other player. So some people just want to maximize their total number of points, ignoring how much the other player gets. Others appear to be trying to match the other player’s score exactly, maximizing their own score while still not “losing.” Finally, others seem to be concerned only with “winning,” always choosing the smaller amount of points but always getting more points than the competition.
But when the responses were broken down according to who was being paid real money, the pattern changed dramatically:
If real money was at stake, just 3 out of 100 participants chose the smaller number of points every time. Most participants either went for the 50-50 option or always chose more points for themselves. But in the purely hypothetical game, where points couldn’t be cashed in, almost no one attempted to maximize their total points — they went for the win instead. This has very important consequences for researchers trying to save grant funds: people behave differently when they’re motivated by real money as opposed to hypothetical money or points. The low-budget experiment that doesn’t actually offer participants real rewards is not equivalent to one where the subjects are actually compensated based on their choices.
But there was no difference in the results when the students believed they were playing against a real person or a computer. Even though there was no interaction between the computer and the study participants, they still tried just as hard to “win” the game, especially when winning didn’t cost real money.
In a second experiment, Kennelly and Fantino changed one key element. Instead of referring to the choices as a “game” with “players,” they presented a new set of instructions about “scenarios” with “participants.” In this experiment, the results were very different. Regardless of whether real money was at stake, only 4 of 156 participants chose fewer points (and therefore more points than the other “participant”) every time. When real money was awarded, nearly all the students chose to optimize their personal reward, and when they were only playing hypothetically, nearly everyone chose to equalize the amounts they earned. Again, there was no difference between a computer and human “opponent.”
In Bazerman et al.’s 1992 study, a different result was found. Even with hypothetical money, participants chose to maximize their own earnings, rather than to equalize with a “competitor.” How can the two studies be reconciled? Obviously Bazerman’s team was using a very different scale of earnings: the maximum amount it was possible to earn in Kennelly and Fantino’s experiments was just over $10, while in the older study the reward could be as much as $600. It’s also possible that Bazerman’s business students had different attitudes about money than Kennelly and Fantino’s psych majors. In fact, our little demo may be able to shed some light on that issue. If our readers choose the $500/$500 split rather than the $600/$800 split, that suggests that the different participant pools explain the differences between the other studies. Otherwise, some other difference between the two studies is probably responsible.
Either way, it’s quite clear that the framing around these questions is almost as important as the questions themselves. Even when real money is involved, people will behave very differently depending on how they’re asked to choose how to divide money between themselves and others.
Kennelly, A., Fantino, E. (2007). The sharing game: Fairness in resource allocation as a function of incentive, gender, and recipient types. Judgment and Decision Making, 2(3), 204-216.