School Signal Investing

Tyler Cowen and Bryan Caplan have been arguing about the signaling theory of education. Bryan’s latest is here, Tyler’s here. Tyler cites two recent empirical papers which estimate that “signaling accounts for one-third of the educational wage premium.” Such papers generally try to predict later income as a function of schooling years and quality, plus other covariates. Today I want point out two bits of “armchair econ” data that complicate such discussions.

Datum #1: Pretty much no one cared about the grades I got before tenth grade.

As long as my grades weren’t very low, I was consistently promoted to the next level. If I got into special classes for gifted students, that was based on IQ tests, not grades. Colleges later cared about my high school GPA, but not about any earlier GPAs. So since no one was watching my earlier grades, they couldn’t possibly be signals right?

Not so fast. Assume for the sake of argument that employers want high GPA grads of good colleges entirely for signaling reasons – students never learn anything useful for future jobs, but merely show innate abilities. But also assume that it takes years of training and effort for kids with high innate ability to learn to do well in school. Given these assumptions third grade schooling is entirely an investment in a signal to later show employers. GPAs then only help kids see how well they are learning school stuff, which will later help them send a good signal. So even though no signaling is happening in third grade, everything third grade students do might be an investment in later signals.

Datum #2: When I visit private firms, people there often mention the prestigious degrees they have.

Yes, they may do this more often for academic visitors, but the point remains. Firms want to impress customers, suppliers, investors, etc. with the quality of their employees, and hiring graduates from prestigious schools helps them signal such quality. Hiring such graduates can also help a manager to impress his bosses, potential employees, and sister divisions about the quality of his employees. Thus even once a boss has determined the “real” productivity of his or her employer, he or she should still be willing to pay extra for employees with prestigious degrees.

Furthermore, people in business signal to each other all the time. In fact, most of the on-the-job business learning that employees do after college, such as how to dress well, how to give presentations, how to write memos, how to talk with clients, etc. might be skills that are mainly useful to signal innate features to bosses, co-workers, clients, etc. So employers might pay more for students with prestigious degrees because such degrees show an ability to learn how to later send good business signals. And this extra pay for top degrees could be entirely an investment in signaling, even if after hiring someone no one ever knew of or mentioned their degrees, and even if schooling makes students better able to please employers.

Bottom line: If much of human interaction is signaling, then much of human investment is in ways to better signal. Businesses that signal are also willing to invest in better signals. The fact that attending school seem to cause changes in students that employers are willing to pay for does not show that school isn’t all about signaling.

GD Star Rating
Tagged as: , ,
Trackback URL:
  • Dave

    My epiphany on education-as-signalling came this morning. If education was adding value beyond signalling, why would it be so tightly age-stratified? Surely there’s nearly as much value in increasing your productive effectiveness at age 40 as there is at age 18, particularly given how quickly patterns of production change nowadays. If it were valued for somewhat nebulous “improving your perspective” reasons, 40 year-olds would probably see more value in it than 18-year olds. And yet, you’re vastly more likely to see 18 year-olds purchasing education than 40 year-olds, probably by a couple orders of magnitude. This makes sense if education is valued for signalling and network effects (including most especially signalling and network effects for mating), but not much if education were valued for content.

    • Captain Oblivious

      Of course, it’s also possible that most 40-year-olds have families, mortgages, jobs, etc, which make it relatively impractical to drop everything for 4 years and go back to school.

      Oh, and I suspect that the 18-year-old’s parents purchase more education than the 18-year-olds themselves do.

      • Captain Oblivious

        OTOH, much of the reason the parents are willing to spend so much is probably to signal that they have the money and good genes (or at least good parenting skills) and good connections to get their kids into top-notch schools.

      • Of course, it’s also possible that most 40-year-olds have families, mortgages, jobs, etc, which make it relatively impractical to drop everything for 4 years and go back to school.

        They can’t afford to earn more, you mean?

        Less-snarky: If education really fed you stuff that amplified your productivity (mostly independent of signalling), then they could get loans such that the higher earnings would more than make up for it. But 40+-somethings have enough signalling built up for it usually not to be worth it.

      • Captain Oblivious

        Seriously? I make $125K per year, and spend most of it (as sole breadwinner) to support my family. You’re proposing that I go to the bank and ask for a half-million-dollar loan, because I plan to quit my job for 4 years and go back to school.

        I think the loan officer would roll his eyes into the next STATE.

      • Right, Captain_Oblivious, because the gain in salary wouldn’t be enough to cover the loan costs. Because the signalling theory is true, and the education would not provide more signalling than you already have. Which was my point.

    • FiftyTwo

      Isn’t the expected value of a degree higher if it is acquired young? Assuming a productive life of 70 years and a four year degree, an 18 year old can expect 48 years of benefits from a degree, whereas a 40 year old can only expect 26. Making it a more sensible decision to apply as early as possible.

      [This is amplified in careers like doctors and lawyers where the early wages are low but wages at 25~ years in become significantly higher.]

  • Eric Falkenstein

    I work at a small firm and most of us are pretty smart. Grades, degrees, are not very important, though there are lots of good ones (eg, Caltech Phds). Being small and all pretty smart, we are pretty indifferent to credentials (conditional upon getting in, which is helped by credentials to some degree). I think getting into a position where your colleagues judge you w/o credentials, and feel appreciated is a good path to satisfaction, because it’s an authentic appreciation, not something manufactured, so its more durable.

    • blink

      Internally, this sounds right: competence and personality matter most. Still, as Robin points out, don’t you feel inclined to mention the credentials of your colleagues when talking to customers (or posting a blog comment)?

  • GregS

    Check out the actuarial career path if you want to see an intense example of post-college signalling. We have to take a series of about ten exams (depending on how you count them) to gain the top designation (“fellow”). Most of the syllabus on the exams is worthless stuff that either will never be used or could easily be referenced if it were needed. I would insist that the signalling accomplishes something, though. I’ve noticed that the people who quit after taking two or three exams are of a different caliber than people who continue taking exams all the way to the end. It definitely takes a certain work ethic (and some innate intelligence maybe?) to pass five or six exams.

    I think there is definitely something to the idea that managers want a lot of highly-certified experts working under them. High-level employees confer some status to their bosses. I can see that in my profession.

  • Drewfus

    Firms want to impress customers, supplies, investors, etc. with the quality of their employees…

    I think the purpose of signalling is misunderstood. It’s not about showing off or using credentials as a proxy for character or whatever, rather it’s a method of arbitrating for expertise in fields that feature significant levels of uncertainty, wrt scientific knowledge. The educational qualifications are simply a way dealing with knowledge uncertainty, by creating an artificial hierarchy based on academic credentials, and basing decisions on the advice of people, justified by their position within this hierarchy, as opposed to references to scientifically based prediction or demonstration.

    Signalling is very much related to the “know-it-all” culture of the West, a culture that appoints people to positions of expert, regardless of the fields general scientific reputation, or the personal record of achievement of the said expert.

    By rephrasing your quote, I mean;

    Firms want to justify themselves to their customers, supplies, investors, etc. with the (semi-arbitrary) status of their employees

    Market signalling is a means of justifying the status of individuals, departments and orgs, not a inefficient proxy for favored qualities. Get over the Peacocks tail explanation!

    • Your explanation fits within the range of signaling explanations I find plausible. In your case “quality” has a socially constructed element to it.

      • Drewfus

        I see signalling as more about status defaults than the communication for impression per se. The signals aren’t there to tell us who is best, rather it’s a socially pre-defined method of regarding people as better or worse as the case may be, when the “weight of evidence” is poor/ambiguous.

        Why don’t pundits or their competitors/critics collect track records of their predictions? Could it be that we (society) do not want to contaminate or complicate our status defaults with actual data that could contradict and therefore undermine the pre-conceived heirarchies of authority we have developed? This seems to have a lot to do with dealing with uncertainty and remaining confident, something that is critical in any society.

        So, yes, quality defined socially (pseudo-quality to some extent).

        Another take on this might be that we like to send positive signals to avoid being discriminated against in disorderly/random/uncertain environments – based on this paper.

  • Torben

    What would be the ideal test (not considering real-world ethical limitations etc.) as to how much education is for signalling reasons versus training?

    • Doug S.

      Give a group of Harvard students answer keys to all the exams (and free essays, etc.), and see if they’re as successful later on as people who actually did the work?

  • B

    First your post about abandoning the the internet for a million dollars shows how exceedingly difficult it is to design a survey. The conclusion I got is ask the question 20 different ways to tease out all of the cognitive biases.

    Now this one makes it nearly impossible to separate signaling from investment from investing in future signaling.

    I’m switching to physics. At least we’re reasonably confident there aren’t infinite layers of complexity there. Just a few more particles to go and we’re done.


  • “The fact that attending school seem to cause changes in students that employers are willing to pay for does not show that school isn’t all about signaling.”

    Agreed. Investment in schooling is a necessary but not sufficient for signalling. Signalling is one of the most visible positive externality of investment in schooling that could be internalized as human capital for capitalization. Exactly because it is the most visible positive consequence for investment in education, society/market turn to downplay other positive externality of education (self-emancipation, scientific advancement, social stability, etc.) and establish this simple quasi cause-effect relation between education and signalling.

  • Gloria, Retired Prof

    A craftsman can show you what’s in his mind by making a material object. A new manager or other office-based worker can only show what’s in his mind by naming the institution he got his degree from. That university normally has a kind of “known” curriculum. I have a very good idea of what an MIT graduate is likely to know and be skilled at doing and have a different idea about a Yale graduate and yet a different idea about a U of Florida graduate.

    The “known” curriculum, often a stereotype, tells me about the aspirations of the graduate and how much work he is willing to do to achieve those aspirations.

    The current degrading of the university due to grade inflation, admissions based on factors other than merit, and the excessive intake of formerly non-admissible candidates is leading to a loss of the signaling effect.

  • Jason

    Adding to Datum #2, I have noticed my management frequently inserting the “Dr.” in front of my name on the title pages of presentations where I had left it off.

  • Pingback: Signaling your ability to signal — Marginal Revolution()

  • Buck Farmer

    Working at a fairly conservative multi-national firm, I can say that signalling comes into play at multiple stages:

    1. During recruiting, it is easier for Recruiters / HR to win management approval by limiting the scope of schools they go to / accept from to a few prestigious institutions. When I’ve seen a push from management to get “better talent” HR doesn’t significantly revise its interview or evaluation method, they shift where they do recruiting.

    2. On the job, there are very strong unspoken social norms…arguably much stronger than any I’d encountered during school. These govern dress, manner of address, who to e-mail how and when, what kind of questions are acceptable and which are embarassing, etc. Partially these norms exist to continually signal solidarity with the group, partially they exist to filter out high vs. low social intelligence.

    Something I haven’t figured out yet is why the social norms in the workplace seem so much stricter than in any other environment I’ve lived in. I suspect it has something to do with multiple generations living together (might be something here for Robin’s forager vs. farmer model).

  • Two things that even the richest person cannot control:
    – the amount of time available to them in a day.
    – the complexity their able is able to absorb and process.

    Many behaviors common in business are not just signalling of ability, but also directly meant to address these limitations of other people. A neat desktop decreases the information processing needed to evaluate that desktop by someone who glances at it, and allows that person to avoid remembering the possible implications of a messy desktop as they continue in their day.

    Now, one might say it’s more logical to 70 minutes doing a job, than to spend 60 minutes of your own time and $200 for someone to reduce the time needed to complete the task by 10 minutes. But there are some things that money can’t buy, and one of those is esteem of someone who makes a decision on whether to associate with someone based on if they have the kind of income where they can afford to spend $200 to save 10 minutes.

    If the demand is for being illogical, then being illogical is logical. And in the end, a lot of this is based on an environment which teaches that any increase in a country’s GDP is positive for that country, without worrying about how that money is being earned or spent. People are, after all, only morally responsible for the information they are exposed to and able to process, and much of the information people encounter is being processed in a flawed way. Thus the criticism of the effort to balance the budget by asserting that it is somehow based off an understanding that it will lead to short-term benefits, when in fact the effort to balance the budget is based off long-term benefits, and if people are interested in short-term improvements they need to re-evaluate what is in front of their nose about rich people not spending their money and the way rich people are getting money at all, and support a reduction in the average wages of full-time work with the option of working part-time at the original wage so that people at profitable companies work less and spend more carefully, thus reducing income inequality and unemployment by diverting purchases to companies in the middle of the quality spectrum with no increase in government spending.

    • >- the complexity their able is able to absorb and process.

      their mind is able to absorb and process, they’re able to absorb and process,

    • Buck Farmer

      – the complexity their able is able to absorb and process.

      This is a perennial bugaboo for me.

      I am told every day about how little time executives have, and consequently how little data and nuance we can give them. Frankly, based on shaped and manipulative most of the proposals I see going to senior management I’d be surprised if decision-making in firms was in anyway a productive process for finding the correct course of action.

      Honestly, I think most senior management should:
      1. Develop or be replaced with people with higher processing power / RAM so they can absorb and synthesize the information necessary to make a decision.
      2. Design the institutions that surround them to obviate their role as a central processor i.e. prediction markets would be one example.

      Instead I see:
      1. Tremendous amounts of wasted work manipulating and obfuscating the flow of information to support short-term local self-interest.
      2. Social norms around dress, demeanor, personal space, personal interests, conversation et cetera that are far more prescriptive and detailed than high school.

      It is not for nothing that I became convinced in my first month working that I was in a modern day Versailles.

      It is sad to see adults begin to indulge more and more the worse habits of schoolyard children while controlling and influencing the lives of millions and billions of people.

      • Buck Farmer

        Ugh. That post was riddled with style and grammar errors. My apologies for confused or thoughtful readers:

        Frankly, based on how shaped and manipulative most of the proposals I see


        Design the institutions that surround them to obviate their role as a central processor i.e. e.g. prediction markets would be one example.

        Venomous anger does not a careful writer make.

      • >1. Tremendous amounts of wasted work manipulating and obfuscating the flow of information to support short-term local self-interest.

        If not for this tremendous amount of wasted work, unemployment would be much higher than it is :P…

        Which again is why the concept of wages needs to be updated, so wasted work does not need to exist like this; and since so many people in the US believe that there are no jobs because companies are doing poor, this is accomplished with a decrease in marginal wage at full-time work, rather than in increase in marginal wage for part-time work.

  • “Tyler cites two recent empirical papers which estimate that “signaling accounts for one-third of the educational wage premium.” ”

    Note that the countersignalling which can arise from an inaccurate signal was not mentioned in the below comment on this estimate:

    Here is the Hanming Fang paper (IER): “…productivity enhancement accounts for close to two-thirds of the college wage premium.”

    The assumptions of the model limit its applicability to the real world, of course. The model assumes that ideal compensation for linear increases in productivity are also linear; or in other words, that someone without the signal or the education could provide the same functions to a business, only at a lower rate, that all products are being sold to consumers with the same marginal utility for money. However, this does not match the pattern for purchases made in the real world, where there are different qualities of goods and services, not just different quantities, and the increase of objective utility of objects with their price is by no means in proportion to the “production efficiency” needed to create them. In other words, “signalling by consumers distorts the payoff for signalling by employees”.

    But generally speaking, it does benefit a society to have signals, if those signals are accurate and measure what people expect them to. A signal will have decreased or even negative benefit to society if it becomes too inaccurate, so one of the objectives of a society should be to find and retain quality signalling methods.

    More specific weaknesses of the model:
    – I assume that each agent has two characteristics (a, v) wherea ∈ A is his innate ability in efficiency units and v ∈ V is his utility cost of attending
    college in monetary terms. weakness: there are costs that cannot be made up for with a higher wage after the completion of college.

    – It is assumed that an agent privately observes his characteristics, but it is commonly known that in the population (a, v) is distributed according to a joint c.d.f. G. weakness: measuring systems have utility for the individual apart from signalling, because the individual often does not know their own characteristics. The distribution of performance is also often not commonly known in a population.

    – Firms behave competitively in the labor market. weakness: many companies have a specific culture of “values” and try to hire people that will fit in with the culture, such as when companies selectively hiring people that graduated from the same educational institution as current employees; in other words, non-financial motives in both hiring and product placement decisions.

    Most importantly, “Stage 3: Test signal. The firms observe a noisy but informative signal, θ ∈ [0, 1], about whether or not the agent is skilled.” weakness: in the model used, skill is acquired at additional cost independent of general efficiency progress from education. Many employers do not test at all for skills for people who are unable to send a signal, heavily distorting the rewards for signalling. In other words significant problems arise when high school education does not even result in a wage offer of minimum wage.

    In general, estimation of the value of a signal becomes difficult because it depends on both the goals of potential transmitters of signals, and also the progress that is represented in the signal’s measurement which will change over time as when multiple standards are all purported to represent the same signal; this causes changes in the distribution of competence for both those signalling and those who choose not to signal. This change in the quality of a signal over time is, perhaps, one of the larger problems with trying to quantify the value of a signal such as in this paper.

  • The 3rd grade performance signals to your parents. Parents worry if they do not get good signals early about their kid’s chances of making a big success later.
    BTW parents signal their friends through those bumper stickers and other ways. Having good performing kids is a signal that I am a smart goo parent.

    • Buck Farmer

      Does anyone know if parents with high-performing kids are more likely to have extra-marital affairs?


      High-performing kids signal that parents can produce high-performing offspring to other adults.

      If we divide positive signals roughly into categories of “attracting mates” and “attracting alies,” I’m pretty sure this would fall in the former.

      So these parents should have more and higher-quality options for extra-marital affairs.

      Anyone got data?

  • Pingback: Overcoming Bias : How Many Levels?()

  • Pingback: Todellisuuden omistaja » Taiteen tukeminen()

  • Pingback: Is higher education the next bubble? - Quora()

  • Pingback: Overcoming Bias : Suits Show Signal Scope()

  • Pingback: Overcoming Bias : How Social Are Signals?()

  • Pingback: Food for thought: week 9 » WHAT I BELIEVE()

  • Pingback: Schools and Signals « azmytheconomics()