« Assorted links | Main | How did the tax cut get so big? »
Do we compete more against fewer competitors?
This caught my eye:
If you've ever had to take a test in a room with a lot of people, you may be able to relate to this study: The more people you're competing against, it turns out, the less motivated and competitive you are. Psychologists observed this pattern across several different situations. Students taking standardized tests in more crowded venues got lower scores. Students asked to complete a short general-knowledge test as fast as possible to win a prize if they were in the fastest 20 percent completed it faster if they were told that they were competing against 10 people rather than 100. Students asked how fast they would run in a race for a $1,000 prize if they finished in the top 10 percent said they would run faster in a race against 50 people rather than 500. Similarly, students contemplating a job interview or Facebook-friending contest said they would be less competitive if they expected more competitors - even if "winning" only required finishing in the top 20 percent. The authors conclude that competitiveness was curtailed because the larger the group, the more difficult it is to compare oneself directly to others.
The original paper is here, but note that context effects may well give you varying results in other settings. The initial article, from the Boston Globe, discusses several other social science mechanisms of interest, although I was not surprised to learn that your dog relaxes you.
Addendum: Here is my earlier article on invisible competition.
Posted by Tyler Cowen on January 5, 2009 at 01:31 PM in Science | Permalink
Comments
This is definitely true for me, in my undergraduate classes.
Posted by: Zachary Kurtz at Jan 5, 2009 1:40:20 PM
I was actually thinking about this for some reason the other day. I wonder if it's because we compete only if we have a chance of winning. Why compete just for the sake of competition, if there's not a chance at glory? If more people are competing, there is less chance of victory. Likewise, if you're competing against someone way out of your league, you're likely to stop competing as well.
Posted by: Chuck at Jan 5, 2009 1:48:13 PM
I'd offer an alternate hypothesis for this phenomenon: That in smaller groups, we care more about our relative rank than in larger, more-depersonalized groups. And "depersonalization" may occur only by knowing the size of the group within which you're competing - being ranked 9/10 might be considered worse than 90/100th, for example. This would be relatively easy to test.
Posted by: Peter Twieg at Jan 5, 2009 1:56:07 PM
These results do not strike me as inconsistent with a standard Bayesian utility maximizer. Take the test example, most people must realize that they could never be in the 20% of all test takers. Their only hope is that the sample is full of stupid people. This is much more probable with a group of size 10 than it is in a group of size 100, therefore it is more worthwhile to exert more effort.
Posted by: Michael at Jan 5, 2009 1:56:13 PM
It all comes back to incentives in my mind. If a test is curved, then I'm definitely going to want to outperform all my competitors in the lecture hall during the exam, but if it's not scaled, then I want to to only take on the marginal costs necessary for the % I want on the exam.
Maybe there's even some sort of Nash Equilibria junk related (I'm admittedly a 2nd year undergrad and know nothing of game theory) on how students during a curved exam compared to when on an unscaled test they have a certain score or % they wish to attain and know the abstract costs involved.
Posted by: Economic Geography at Jan 5, 2009 2:03:45 PM
This behavior could be rational.
Say you think you're in the 30th percentile and only the top 20% get prizes. In a smaller group of people you are more likely to get lucky and outperform your estimate (maybe your competitors are worse than the general population) and you should therefore compete harder. But the more people you are competing against the closer your outcome will be to your true skill level.
So people in the top 20% should always compete hard. People near the bottom should never compete hard. People at the margin should adjust their efforts based on population size.
Posted by: Brian at Jan 5, 2009 2:07:36 PM
Why didn't the authors conclude people compete less intensely because it's less likely you'll win in a larger group? Larger samples mean less delta between the sample and population distributions, so regardless of your absolute skill level (assuming the contest is skill-based in some way, and you are not near the extreme high end) as the number of competitors goes up, the probability you will be able to rank in a given percentile goes down.
Posted by: Noah Yetter at Jan 5, 2009 2:07:44 PM
While I'm certain the effect is mostly psychological, variance is inversely proportional to sample size. So you should be more assured of your relative position in a larger sample size than in a smaller sample size.
Posted by: Jody at Jan 5, 2009 2:17:07 PM
"your dog relaxes you" yeah, but what about him ?
Posted by: Hans at Jan 5, 2009 2:17:11 PM
I'm applying to graduate schools now and I have little idea what the competition is like for these spots. It is harder to do them for the reason stated in this paper. It is stressing to work on applications when you don't know whether you're even competitive or not, especially since there are other things I'd rather be doing if this won't payoff. (In reality, I think I will get in somewhere, but I don't know at what school and at what level, so I am applying to A LOT of programs.)
Posted by: Robert S at Jan 5, 2009 3:15:57 PM
I agree with Michael and Brian, but would argue that it works the other way, too. If I know that I'm in the 90th percentile, I can just go through the motions in a large group and know I'll get a prize. But in a small group, there might be a couple of geniuses in there, and I may lose if I don't work as hard.
Posted by: Phil at Jan 5, 2009 3:38:13 PM
Isn't there a difference between smaller and larger groups that isn't being accounted for? In a smaller group, the standard error of scores is higher (because the se metric depends on sample size).
For example, if you believe that you are at the top 20th percentile of the general populace, your chances of falling into the top 10% of a given sample are much higher if that sample size is smaller. More specifically, with a larger sample size, your chances of falling into a real percentile above your expected percentile -> 0 as your sample size -> the population size. Or did I miss something?
Posted by: blake at Jan 5, 2009 3:51:27 PM
50% of the people have absolutely no idea what you are talking about when you tell them "they need to beat 20% of the people."...percentages completely baffle them, thus they ignorantly say they will try harder if there are fewer in the test population.
Posted by: Gabe at Jan 5, 2009 5:07:00 PM
When I was in college (classes usually smaller than 20), I met a guy who always got the best grade in our joint classes.
When I asked him how he did it, he told me he was always competing. With himself.
A group of 1.
Posted by: at Jan 5, 2009 5:50:37 PM
Is it reasonable to imagine that being able to better personify and quantify your competition makes you more determined to beat it?
My company has only two real competitors; GE and Dresser Rand. When we are not dwelling on the fact that GE could eat us with their petty cash, we work harder to beat them in X statistic. (profitability, quality of machine, outbid on a certain project...) if we had more than 2 competitors, it would be harder to get worked up about things. This is why propaganda reduces complex issues to a few salient (and usually insufficient) points. It is easier to motivate people when they have a definite objective or enemy.
Posted by: Sean at Jan 5, 2009 10:58:35 PM
In a smaller population individual effort will amount to more and thus be worth more effort so this makes sense.
Posted by: Lord at Jan 5, 2009 11:50:17 PM
This probably has nothing at all to do with the post, but I did find to my surprise in law school that the bigger the class, the better I did. I regularly got the highest grade in classes up to 150, but in classes of 5-15 I did absolutely terribly (relatively speaking- about 1 point of GPA lower). Very puzzling. Well, the classes probably were qualitatively different, but even so.
Posted by: Cliff at Jan 5, 2009 11:52:23 PM
Perhaps the per person marginal "competition" falls while the total "competition" for the group rises. Consider very large groups such as world equity markets.
Posted by: rluser at Jan 5, 2009 11:56:30 PM
i find it immeasureably easier to get into the top 700 of a 3500-person poker tournament than the top 2 of a 10 person table as there are many more stupid people willing to give me their chips while i avoid those better than me
Posted by: rtgkjetr at Jan 6, 2009 2:11:56 AM
I would more be interested in the behaviour of some specific individuals at certain points, rather than a description of the aggregate behaviour of the group. For instance, how more work do the people who have a good chance at finishing at the very top or very bottom do? I'd imagine that people compete more to finish at the top of a 100 person group than a 10 person group, all other things being equal. However, since a relatively smaller proportion of the 100 person group has a reasonable chance of winning, fewer will make an attempt. The increasing marginal costs of additional effort mean that even with the same expected prize there will be less work.
Posted by: Nichlemn at Jan 6, 2009 6:54:09 AM
Check out contest and tournament theory in labor economics.
Posted by: bob tollison at Jan 6, 2009 7:44:57 AM
There are too many comments in this thread for me to bother to try to write one that stands out.
Posted by: Yancey Ward at Jan 6, 2009 11:52:01 AM
@Bob Tollison:
Right, I was just going to say that. How is this different from the result one gets in rank-order tournaments, where people will provide much less effort if they expect to be far from the cutoff? This doesn't strike me as a really new finding...
Posted by: Lupin at Jan 6, 2009 12:06:37 PM
Ofcourse it would follow then that a good strategy when competing against a large group is to isolate a much smaller sub-group of friends (or foes?) to compete with. Members of such a group would increase their competitiveness compared to other individuals in the larger group.
... unless perhaps those other individuals become aware of the new sub-group. Then the new sub-group could become the focusing point for all of the other individuals, fueling their competitiveness?
The more tangible the opponent, the easier it is to compete?
That seems to be the conclusion of the researchers anyway, as the article notes "the larger the group, the more difficult it is to compare oneself directly to others."
I wonder then what would happen if the researchers had given a detailed generalization of the individuals making up the rest of the group? Would people compete more against a group of 100 blond swedish female supermodels in their 20s vs just a group of 100 random people?
Posted by: Mike at Jan 6, 2009 11:01:20 PM
Why must we rely on others to determne our own personal success? Why cant people single themselves out of the crowd and act as if they are the only people there? This may sound somewhat delusional and emotionless, but it really works. I earned good grades throughout my first few semesters in college by maintaining that train of thought. Everyone becomes sort of insignificant in your determining of future actions. Its all about the self and discipline. Its the same kind of feeling you get when someone challenges you at something youre realy really good at and you know youre gonna win. That same feeling must occur without having that external challenge. Nothing should stop you.
Posted by: Nicola Mushahwar at May 21, 2009 2:21:38 AM
