Arnold Kling  

Signaling and Costs

PRINT
Signaling and Vicky Clubs... Lessons from Solyndra...

Bryan writes,


when you make signaling cheaper, agents' natural response is to signal more intensely or on another dimension.

My understanding of the signaling model is that it depends crucially on the relative cost of signaling to people with and without the desired trait. You want the cost to be high for someone without the trait and low for someone with the trait.

With that in mind, I do not see how lowering the cost of signaling for people with the trait does anything other than cause people with the trait to choose the low-cost signal. The problem with a low-cost substitute for a diamond is that it lowers the cost of signaling for people without the desired trait (which is a willingness to buy an expensive gift).

If I come up with a low-cost way to earn a badge that signals intelligence, conscientiousness, and conformity, and that badge can only be earned by people with those traits, then my badge should find a market. One challenge is that when few people use the badge, it seems to signal non-conformity. Thus, the early adopters of my cheaper badge do not do as well as they should. But over time, there are two possibilities. One is that the conformity hurdle cannot be overcome, so that the incumbent signaling mechanism remains dominant forever. The other possibility is that eventually a tipping point is reached, and enough people use the new badge so that it no longer signals nonconformity. At that point, the market position of the old badge rapidly deteriorates.

I think that we will arrive at the second equilibrium at some point. However, predicting when it will occur is difficult.



COMMENTS (7 to date)
blink writes:

A tipping point is plausible, and for intelligence and conscientiousness I think you make a good case that we may find cheaper signals. For conformity, however, the bar is much higher. How can a signal that initially indicates anti-conformity come to signal conformity? You mention this problem, but I am not convinced that it will be overcome.

J Storrs Hall writes:

OK, a gedankenexperiment:

Suppose there is a mechanical educator, a la Asimov, such that the cost and time of getting a college education is the same as a trip to the dentist. Let's imagine that besides educating you, the machine prints out a certificate saying how educated you were (100=Harvard PhD, 0=Pleasant Valley Community College dropout). Assume the certificate reflected the actual result at least as well as the current real degrees do.

How much difference would it make to an employer whether everyone going in to the educator paid $250, or you paid on a sliding scale that was 1000 times as pricey for the top certificate?

Second question: suppose the machine could be used in a purely analytic mode, without educating, but accurately determine your intelligence and conscientiousness, etc. I claim it would be illegal to use it in hiring for the same reason as IQ tests are: disparate impact.

This forms an ill-defined lower bound on the price of such a filter, namely that it has to look enough like something that isn't a filter.

Ken B writes:

Arnold's criterion predicts good looking people will post photos on craigslist sex ads and ugly people won't, and Bryan's criterion suggests they will flee craigslist. Any bets?

@blink: piercings are now a conformity signal.

Glen Smith writes:

Costs are key to the value of the signal.

Glen S. McGhee writes:

" ... my badge should find a market. One challenge is that when few people use the badge, it seems to signal non-conformity. Thus, the early adopters of my cheaper badge do not do as well as they should. But over time, there are two possibilities."

Not at all. Another possibility is that other uses than those intended will emerge: badges are like AP exam grades of 4 and 5 are for high schoolers. They end up on college applications, and nothing more.

Dan Carroll writes:

The issue with Bryan's argument on the signaling model of the education system is that he is assuming that that model is stable. It is a model that is predicated on the assumption that there are not other ways to signal status (such as working at Goldman Sachs, joining the right club, wearing the right cloths, etc.). Education as signaling depends on perpetuating the myth that one goes to college to learn something and not signal. As long as enough people believe that, as long as the great subsidizer believes that, then the model is vulnerable to disruption - especially at the low end.

Every culture has signaling models, and there is no reason why the model can't evolve, as it has over the past 50 years. Yes, starting a new model from scratch is not a good business venture, but we already have standardized testing systems in place and there may be other alternatives waiting in the wings to take center stage.

Bryan Willman writes:

Life is always graded on a curve, except for passing genes to the next generation, which is pass/fail.

Signaling is interesting as it relates to filtering - if "too many" people have The Signal, it loses its leverage.

First rule of how membership in Elites is Defined by said Elites -> "Not You"

Second rule of how membership is defined - it changes over time, to enforce the First Rule.

So even if the educational robot gives, say, 1 million americans legal skills equivalent to the Chief Justice, virtually all will be unable to fulfill that particular potential.

That one is sort of obvious - but people miss that similar "only X people" rules apply not only to "who can be in the top 1%" but "who can have the most interesting jobs at ...."

So removing the waste of taxpayer and probably some donor funds from inefficient allocation of educational resources is a fine thing, but nobody should think than any amount of tinkering with education will result in life being fair.

Life is graded on a curve, period.

Comments for this entry have been closed
Return to top