Universities and Economic Growthby Philip Greenspun in June 2009
|
Site Home : Teaching : One Article
Value of $150,000 spent at Harvard: I drove to downtown Boston with a
guy who was in between his third and fourth year at Harvard
University. I got out and said "Please park the car near my old
apartment. It has a Cambridge resident sticker on it, so you should be
able to find something by driving around the block of Sumner Road,
Kirkland Street, and Irving Street." Twenty minutes later an instant
message arrived: "I parked your car in a tow-away zone for a moving
van. The only alternative was a fire hydrant." A few hours later, I
asked about the possible alternative of driving one more block and
parking the car legally. "I couldn't do that. You told me to park the
car within one block."
Value of $120,000 spent at Emerson College: I hired a summer intern, who had finished three years studying film and video production at Emerson College, to work with me on a series of videos for people learning to fly helicopters. He brought his own equipment to the first day of filming, including some "work lights" from Home Depot. To get a clear image of the instrument panel, he positioned these lights directly behind my $3000 Sony high-def camcorder. Within a minute, the camcorder screen melted and the entire machine was nearly hot enough to catch fire.
We know the cost of buying another camcorder from Japan ($3000). What's the cost to American society when we have to take an afternoon off work to retrieve a towed-away car? More importantly, can we afford these costs in a declining economy that faces competition from Chinese and Indian firms?
The potential losses to the economy from ineffective university education are larger than the $560 billion spent. An American university graduate who is not dramatically more capable than his low-wage counterparts in poor countries puts the entire U.S. economy at risk of stagnation.
What about the industrial revolution? Doesn't that explain why per-capita living standards in the West increased so rapidly? Clark argues that Malthus was right: population will expand in response to an improvement in resources or technology. The result of an increased food supply will not be an improvement in the average person's standard of living, but an increase in the number of people, all living at the subsistence level.
Evidence that industrial technology alone is not sufficient to generate growth is supplied by all of the poor countries in our 21st Century world. Nearly all of the world's scientific and technical knowledge is available in textbooks, journal articles, patents, Web pages, and other public sources of information. A country that has not invested one dime in research or engineering can, practically for free, tap into the fruits of all of the world's hardest-working nerds. A dusty corner of Africa or Asia might be 10 or 15 years behind Japan, Germany, or the U.S., but should not be 50 years behind.
Clark looks at why India did not wipe out England in the production of textiles. Indian mills had the same machines and the same managers, sent out from England, as their competitors back in England. Yet Indian workers were so unproductive that English and American mills remained competitive. Clark gets the rates of "doffs per hour", in which a person removed a full spindle of yarn from a spinning machine. An Indian worker could do 120 per hour, compared to 460 for a Briton and 770 for an American. The machines and tasks performed were identical.
Clark argues that the more advanced the technology, the greater the wage premium for high quality labor. A small mistake in a preindustrial agricultural process meant a tiny reduction in the quantity of grain delivered. A small mistake on an integrated circuit fabrication line means the scrapping of hundreds of thousands of dollars worth of product. A worker who does not follow checklists consistently is not worth employing in a modern factory, even at a wage of 1 cent per hour. (Nor would you want to give the accident-prone worker a $3000 camcorder...)
What lifted England out of its centuries of torpor? Clark argues that it was a very gradual rise in literacy and a genetic personal tendency towards future-mindedness, both caused by the fact that wealthier men had more surviving children than poor men. Because the economy was stagnant for hundreds of years, the inevitable result was downward mobility. The children of a rich man would become small-scale farmers or artisans. Eventually most of the English were somehow descended from people with a psychological propensity to save and invest rather than leap at instant gratification.
The Democratic Party platform for 2008 and many of its candidates complained that private sector workers have not enjoyed increased real wages since the 1970s. Someone who had just finished read A Farewell to Alms would respond "If American workers aren't better educated than they were in the 1970s, why would you expect their wages to rise?"
An alternative formulation of this question would be "If American workers aren't better educated than Chinese, Indian, and Mexican workers, why would you expect their wages to be higher?" This becomes a problem for politicians because the U.S. government could not fund itself by taxing workers paid at the Mexican level, even if it confiscated 100 percent of their income.
High schools are run by governments and teachers' unions. High schools need not compete for students; nearly every student in a particular region is forced to attend a corresponding high school. Every attempt at reform has been about as successful as getting the Pentagon to stop paying $1000 for toilet seats. Colleges and universities, however, tend to be smaller organizations, and they compete for students, which provides some incentive to improve.
The university incorporated an important quality control mechanism: student associations paid professors according to how well they taught, how many students were attracted to their lectures, and whether they showed up on time.
It made sense for students to show up to lecture and to do their homework. A student's lodging might not have been heated. It might make sense to come to lecture simply to get warm. Students in 1088 had no television, no radio, no Internet, no email, no instant messaging, no mobile phone. A student might come to lecture for entertainment.
What about homework? Students in a pre-technological university would do homework either in the library or at home. Both places lacked television, video games, email, etc.
Teaching technologies developed since 1088:
Improved technology has rendered the traditional university instructional method far less effective. The student has a warm cozy apartment and will find sleeping late an attractive alternative to attending a lecture (or watching Good Morning America). The student sitting in lecture has some sort of device capable of browsing the Web, sending and receiving text messages, supporting games, displaying photos or video to an adjacent student.
Focusing on homework has become much tougher. A modern dorm room has a television, Internet, youtube, instant messaging, email, phone, and video games. The students who get the most out of their four years in college are not those who are most able, but rather those with the best study habits.
No company would rely on this system for getting work done, despite the potential savings in having each employee work from home. Companies spend a fortune in commercial office space rent to create an environment with limited distractions and keep workers there for most of each day.
In the face of massive technological advances, the most significant change that universities have made is removing their only quality control mechanism. Through tenure, the university now guarantees professors pay regardless of effectiveness.
A problem with this approach is that it depends on finding millions of great storytellers.
Consider lectures by Robert Shiller, e.g., Econ 252, Financial Markets: Lecture 1 (alternate location):
f(x) = P^x (1-P)^(n-x) n!/(n-x)!all two minutes are taken up with writing (incorrectly; there is an unbalanced parenthesis) and asking if people can read his handwriting. [Compare to two minutes reading http://en.wikipedia.org/wiki/Binomial_distribution.]
A bit later Shiller presents the moderately scary formula for a Gaussian Distribution with no explanation and says "I assume you're familiar with this". Students at Yale must be very intelligent indeed if they can understand the Binomial Formula and Gaussian Distribution simply by looking at an expression. But if they are so smart and math-nerdy, how do we explain this sequence:
This is one of the most popular courses at one of America's greatest universities.
Bureau of Labor Statistics says that 1.7 million Americans work as college teachers (source). If Yale can't find teachers who can use classroom time effectively, what hope is there for universities with less money and prestige?
University of Phoenix charges less than other private universities, about $12,000 per year in 2009 compared to an average of close to $25,000. University of Phoenix earns a profit, and cannot draw on an endowment. Public universities are relatively cheap, at an average of $6,000 per year, but impose a heavy cost on taxpayers, rely on donations, and generally require students to spend four years out of the full-time traditional 9-5 workforce.
In the Internet age, a student's work is just as easily available to a professor on either coast. Why then would a professor be grading his or her own students? It is an invitation to dishonesty. A student who has learned nothing will not receive an F because the professor doesn't want to admit that his teaching hasn't been compelling and/or effective. Having the professor be simultaneously teacher, coach, and executioner sets up a bad dynamic in which students are afraid to admit weakness and ask for help.
It would cost nothing extra to have teachers at University of Kansas grade University of Massachusetts students and vice versa. The teacher at U. Kansas supplies the U. Mass teacher with the course syllabus and standards and the U. Mass teacher applies the standards without the bias of "these are the students who just sat through some lectures by the world's greatest genius, i.e., me."
[Note that this approach also eliminates bias based on race and sex, assuming the grading is done through a computer system that hides a student's name from the grader.]
Of what use is the ability to sit passively in a lecture? How many jobs are there in the U.S. that pay someone to do something like sit passively in a lecture?
What if the teacher needs to communicate something to all of the students in the class? That's why God invented email and the Web. If a professor needs to summarize the best points made in a lab discussion, that can be done via a broadcast email. If all of the students seem confused about a point that isn't covered well in their textbook, the professor can prepare a Web page on the subject and email it to the students.
Students in 2009 should not be required to exercise super-human self-discipline in ignoring the siren songs of email, Web, IM, Xbox, telephone, etc. A college student should be able to show up to school at 9:00 am, break for lunch at 12:30, return at 1:30 pm, and go home at 7:00 pm, all work for the day completed.
Students should work in an open-office environment, where they can get assistance from others sitting at adjacent desks and from roaming teachers.
This system would require that curriculum flexibility be reduced or that the cherished semester system be scrapped. If all sophomores majoring in, say, Economics, were seated together, they would need to be taking the same courses. An alternative would be to teach one course at a time and group all of the students in that course together for the three or four weeks of an intensive class. After the course ended, the students might be reshuffled. This would interfere with the semester system. Instead of teaching a few hours per week for 13 weeks, a professor would have to work hard for three or four weeks and then would have a month off from teaching.
[The author of this article has some experience with this idea. He founded a one-year post-baccalaureate program in Computer Science. The curriculum was copied from traditional schools such as MIT and Stanford, but each course lasted one month. The 37 students sat together in a cubicle farm and worked on homework and labs together. After ten months, they'd taken the ten core courses of an undergraduate CS degree and 100 percent of the students who started the program were able to perform competently as software developers, applicants to Computer Science graduate school, and in other careers typically chosen by people with bachelor's degrees in CS.]
In the working world, someone who can do good individual work is unlikely to be a critical asset to a company unless he or she is also able to help others do good work, either by contributing to a multi-person project or providing feedback on someone else's work. The graduate of a traditional university curriculum has had almost no experience in reviewing peers' work.
The solution to both problems is to build in substantial peer review. In an engineering school, students should present their designs to other students. If critical flaws are not caught during the design review, that should be reflect badly primarily on the peer reviewers, not on the design engineer. Students should learn to be useful editors of each others' writing. If done through a collaborative word processor, an outside grader would have easy access to the first draft of a paper and the reviewer's comments. The grader would be able to give feedback to the reviewer on the usefulness of the comments.
Consider the typical Computer Science graduate. He (for it is almost always men who are dumb enough to major in this dreary subject) has done a lot of problem sets, tackling small problems that were precisely defined by professors. He has never met a client. He has never been asked to do something that isn't doable in the allotted time. He has probably never written more than one engineering document. He may never have worked on an open-ended problem. The CS graduate comes out prepared to work for an engineer, not to be an engineer.
Surely this problem can be remedied by the first employer, right? In poor countries, there aren't enough jobs for entry-level software engineers for the average CS to have any hope of getting one. Most will become system or network administrators. In developed countries, the companies most eager to hire kids fresh out of college are the largest firms. What the new graduate needs is to tackle 100 percent of a small project, building up his skills at listening to clients, writing specifications, and crafting prototypes. What the new graduate gets is 1 percent of a big project, working on a precisely defined slice handed to him by the real engineer supervising the effort.
Why isn't all of an CS undergrad's education done via open-ended projects? Consider the requirement to take Physics 1 (freshman mechanics). This could easily be replaced by the professor saying "I would like you to build me a simulator for my home pool table." The professor doesn't say what the interface should be or how detailed the simulation must be. It is up to the students, who will organize themselves into groups of 2 or 3, to figure out what is practical to accomplish in a portion of the semester. It will also be up to those students to work through exercises in a standard Physics textbook as a subtask.
By the time our CS student completes college, he will have gone through the cycle of "listen, design, discuss, implement, test, refine, and write up" roughly 100 times. He shouldn't have any trouble going through the cycle for the 101st time at his first job.
I have some personal experience with this approach to teaching in one course at MIT: Software Engineering for Internet Applications (online textbook). I find real customers who need online communities built. Course meeting time is devoted to teams presenting either designs or prototypes and the other teams criticizing. This prepares students to participate in engineering design reviews. If nobody makes a suggestion I consider important, I will make it myself at the end of a review or via email to the class mailing list after class. If there is something that I want everyone to know, I add it to the online textbook or email it to the class.
Occasionally we will use the meeting time for teams to work together (sometimes paired with another team) on a data modeling challenge. This works great in rooms with a lot of blackboards and after 20 minutes we start asking teams to present their work while others criticize and comment.
Most evaluation is done by the clients (usually non-technical), professional interface designers whom I bring in, and professional software engineers whom I bring in as mentors and critics. The final presentation and write-up is evaluated by a team of business executives and venture capitalists; students are told to pitch the project as though they were seeking follow-on funding. Instead of me being the bad guy telling a student his work is substandard, I let the client or outside consultant be the meanie.
How has it worked? The better graduates of the course are at Google, Microsoft, or running their own startup companies.
I was asked by Neumont University, a startup for-profit computer science school in Utah, for advice on how to structure the school. They didn't know how unworkable these ideas were so they adopted most of them. Because their student body consists of kids from middle class families, there is no need for a long summer break for students to join their parents on a grand tour of Europe. Nor need the students take a month off in the winter to yacht around the Caribbean. Simply by being in session Monday through Friday, 8-5, for most of the year, Neumont is able to graduate CS majors in about 2.5 years. That's 1.5 years in which the kid is not coming home to ask his parents for more money.
Neumont also adopted the idea of making most learning project-based. Neumont freshmen start with substantially lower SAT scores and high school grades than University of Utah freshmen. With brighter and better-prepared students plus a 150-year headstart, how does U. of Utah's faculty do compared to Neumont's? The graduate of U of U's traditional lecture-and-homework CS program will be 1.5 years older than a Neumont graduate, start at a lower salary, and have fewer job offers.
Improving university education is one of the cheapest and simplest things that we could do to improve the U.S. economy's long-term prospects. The current system, in light of both the promise and distractions of modern technology, is almost laughably poorly designed. A well-designed system ensures that every student who shows up on a regular basis will become competent.
The focus in universities appears to have shifted from education, to living the good life for four years. Universities these days have five-star dorms that cater to students' every whim (double decaf mocha latte at midnight, anyone?). How about we return to modest buildings and dorms (clean, but very basic accommodations) so the money can actually be spent on education? How about cost containment? Surely professors' salaries are not rising at 2-3 times the rate of inflation that tuition is rising at?
-- Jagadeesh Venugopal, October 15, 2009
I'm not sure about an open office. I could study fine in my dorm room, which didn't have a computer or anything like that but in an office I'd have to deal with other people. Important at work but it wouldn't help me get my Chemistry book read.U tended to like studying on my bed while swishing my legs around. Lying on the floor swishing your legs wouldn't be acceptable in an office.
That's not to say I wouldn't have benefited from more collaborative projects, but I need time alone to.
However, I agree about lectures being mostly useless. I often drew during lectures to stay awake.
-- Elizabeth Porco, November 2, 2009
The first day of Statistics 100 at the University of Michigan, the professor said our final grade was either the average score of our midterm and final, or the score of the final -- whichever was greater. Our labs and homework assignments did not impact our course grade. My roommate, also enrolled in the same course, didn't make the first class. I told him about the grading system. Neither of us attended another class, nor took the midterm.Before the final, we spent two days straight, reading the material and doing problem sets. I got a B+. He got an A-.
I am ashamed to admit that I wasted all those lectures and labs that my family paid for. But what does it say about inadequacy and inefficiency of the lecture system when we arguably learned the practical application of our course material with 24 hours of self-study spread over two days and a $100 textbook?
Compare that with the cost of the lectures and labs -- 64 hours spread over 16 weeks and $3,861-5,791 in tuition (depends on course load, based on current tuition of $17,374).
-- Mike Lin, November 9, 2009