Thoughts on business, engineering and higher education

PhD Comics

Today's post is for my readers who need a bit of cheers to start the week, especially those in academia or graduate school. The links below all redirect to PhD Comics, by Jorge Cham.

• Film Crew: Part 1, Part 2, Part 3, Part 4, Part 5, Part 6, Part 7
• The Science News Cycle
• Research topics guaranteed to be picked by the news media.

And also:

• The Underlying Question
• I'm Sure of It
• (Read the following together) Budget Cuts, Humanities Hearing, Boorish Barbarians, Price Tag on the Human Soul.

Enjoy!

Real Life and Formulas

NPR ran an interesting feature recently, criticizing the use of the Body Mass Index to evaluate Americans' health and determine medical insurance premiums (Top 10 reasons why the BMI is bogus, July 4, 2009). Many of the reasons mentioned in the article reminded me of caveats I have heard in other contexts, especially regarding the current economic meltdown. Here are a few:

1. The person who created the formula (two centuries ago!) "said explicitly it could not and should not be used to indicate the level of fatness in an individual." This echoes David Li's warnings about placing too much trust in the models he had pioneered to price credit-debt obligations. (See my old posts: Part 1 and Part 2 of "Finance's Gaussian Copulas", March 2009.)
2. According to the journalist, it has no scientific motivation but was created to fit the data. (The BMI is the ratio of weight to height-squared, times a scaling factor because the formula was initially in metric units. Why the height should be squared is a bit of a mystery.)
3. "It makes no allowance for the relative proportions of bone, muscle and fat in the body. But bone is denser than muscle and twice as dense as fat, so a person with strong bones, good muscle tone and low fat will have a high BMI. Thus, athletes [...] tend to find themselves classified as overweight or even obese."
4. "Because the BMI [...] comes from a mathematical formula, it carries an air of scientific authority. But it's mathematical snake oil."

I'll let you read the rest. The author makes excellent points: it is not enough to have a mathematical formula - it should be built on solid ground. Otherwise you get nothing more than fake "insights". (Also see the second paragraph of this July 2008 post for a related story).

In the meantime, the fascination for mathematical formulas is alive and well, as demonstrated by this New York Times article on the injuries of baseball players (Seeking a Way to Predict Baseball Injuries, July 7, 2009). The Dodgers' director of medical services, Stan Conte, "is trying to build a formula that will give teams a competitive advantage and help them avoid players who spend their days in the training room and not on the field."

He has accumulated large amounts of data over his 15 years as trainer, and "draws on the analytical prowess of David Zes and Adam Sugano, statisticians who teach at U.C.L.A., and Matt Marks, an employee in the Dodgers’ baseball operations department. They are building mathematical formulas that they hope will show the chances a player will be injured within the next season." (Emphasis mine.)

The article mentions a personnel decision for which Conte's analysis ultimately proved correct, and a personnel decision for which it did not - a good reminder that models will never be 100% accurate. It is not clear whether these models will become widely implemented beyond a few teams, although "the formulas could become lucrative if successful". To use an expression found twice in the NYT article, statistical analysis to predict baseball injuries is used by "progressive medical staff"; not all trainers will see value in such mathematical techniques. That could become a new slogan, though: "Be progressive! Use math!"

Girl Power

These two articles from the New York Times are both about a teenage girl outperforming boys in sport, and the resulting tensions:

• "Scary, isn't she?" (September 11, 2008) is about Jaime Nared from Portland, Oregon, who was "kicked off a boys’ basketball team for being too good." ("[T]he last time she played organized ball against girls her age, the final score was 90-7. [Her coach] described the dynamic as “like having Shaq on a high-school team.”") Jaime was banned from competing with the boys after scoring 30 points in one game, which apparently did not endear her to the parents' of the boys on the losing team; gym management suddenly "cited a previously unenforced rule against mixed-gender play." The behavior of the other parents, at one point described as "sexist rage", reminded me of the parents' screams and shouts in a New Yorker's article by Malcolm Gladwell on innovators - the innovator in question was the coach of their opponents, also in basketball, who used a little-known technique called the full-court press and was making their kids lose (see this old post of mine). Jaime Nared then had to play with high school girls, which was not ideal because of the age difference. About a day after the NYT article appeared (including the words "[Jaime's mother] is considering a civil claim"), Jaime was allowed to play with the boys again. 
• "She plays with boys, and rivals don't like it" (June 1, 2009) is about Hannah Berner from New York, NY, who plays tennis on the boys' team because her school has no girls' team and "help[ed] her team sweep the city’s three major high school tennis tournaments this year". ("Thanks to her play and the complaints of some of her opponents, [Ms Berner] may have moved her high school closer to securing approval to field a girls’ tennis team. After Ms. Berner won one recent game, the coach of the opposing team complained that she should not have been allowed to play with the boys because her gender unnerved her opponents." One boy whom she had defeated "shattered his racket in anger.") Hannah Berner will join the University of Wisconsin in the Fall, where she will finally play tennis on the women's team.

Unemployment Numbers

The Local Area Unemployment Program of the Bureau of Labor Statistics offers comprehensive data on employment rates in the US. I wanted to look at the numbers after the latest news on the joblessness rate, which recently reached 9.5% - I was wondering whether the average was driven up by the high unemployment rates in a few states, or whether the unemployment situation was homogeneous throughout the country. (See this old post of mine about issues in using the average to describe the main trend in the data, and the benefits of using the median instead.)

You can find a table of the latest unemployment rates by state including historical highs and lows here; California, Florida, Georgia, Nevada, North Carolina, Oregon, Rhode Island and South Carolina all reached a new historical high in unemployment in May. ("Historical high" means "highest since 1976", which is when the BLS started tracking unemployment numbers. Data for June is not yet available on the site. For comparison purposes with the data by state, note that the national unemployment rate in May was 9.4%, not 9.5%, which is for June.) The BLS also ranks the states by increasing unemployment rates here. This reordering makes it easy to see that the median unemployment rate by state for May was 8.2%, which happens to be the jobless rate of Arizona, Massachusetts, Minnesota, New York and Pennsylvania. (The median is the point in the middle of the data. The table includes the 50 states and the District of Columbia, so the median is the 26th rate, but because of the tie the 26th rate falls in the middle of the 5-way tie.)

Only 17 states and the D.C. had an unemployment rate equal to or higher than the national average - in other words, two thirds of the states have unemployment rates below average. As expected, the (populous) states with high unemployment rates drive the national number up. The states with a rate exceeding 10% are: Illinois, Florida, Indiana, Kentucky, Tennessee, Ohio, North Carolina, Nevada, California, Rhode Island, South Carolina, Oregon, Michigan and also the District of Columbia. Michigan, at 14.1%, has by far the highest unemployment rate.

What puzzled me most about this data is that a handful of states have much lower unemployment than I anticipated. Specifically, Nebraska, North Dakota, South Dakota, Wyoming and Utah all had unemployment rates below 5.5%. I cannot quite understand why those five states are weathering the recession so much better than their neighbors, but this short analysis by a DC think tank offers some answers, although it leaves a lot unexplored. (As an aside, make sure to click on "Play the timeline" on the US map.) Reuters published an article about the low unemployment rate in North Dakota in March, praising its diversified economy and its stable housing market; Standard & Poor's upgraded North Dakota's credit rating around the same time. MainStreet.com ranked North Dakota #1 in its Job Opportunity Index in May. Wyoming and Virginia followed closely behind. (Virginia's unemployment rate in May was 7.1%.) 

More data and charts about the evolution of the US unemployment rate since 1999 are provided by the BLS here; the scale of the upward trend over the past year is sobering. The BLS also provides plenty of unemployment maps for each state by county, for instance for New Jersey and Pennsylvania. I was shocked to learn that Cameron County in Central PA has an unemployment rate of 18.2%. (The county has about 6,000 residents, according to Wikipedia.) Neighbor Elk County, with about 35,000 residents, has the second highest rate of 14.9%. But those numbers pale in comparison to those of California, where many counties have rates exceeding 15%, with even a staggering 26.8% for Imperial County.

Economists surveyed by the Wall Street Journal in April said that "it won't be until the second half of 2010 that the economy recovers enough to bring down unemployment." That is a long time to wait.

Free Online Courses

About two weeks ago, Inside Higher Ed reported that the Obama administration was "in the final stages" of drafting a program to help high schools and community colleges put free courses online, with a price tag of about $50 million a year (U.S. Push for Free Online Courses, June 29, 2009). In addition, "the plan would provide $9 billion over 10 years to help community colleges develop and improve programs related to preparing students for good jobs, and a $10 billion loan fund (at low or no interest) for community college facilities."

The details are sketchy; "[a]dvocates for open courses guess that the proposal reflects the ideas of Martha J. Kanter, the under secretary of education. Kanter was previously chancellor of the Foothill-De Anza Community College District". The Obama administration has declined to comment on the plans until "the time is right", but the program is part of a broader effort to improve the skills of the US workforce, in particular people with no high school or college diploma who typically struggle to find a job. The administration's initiative could help many become more employable.

I do find the scope of the program - not only the free online courses but the rest of the initiative - a bit daunting. The National Skills College in particular sounds like a major undertaking. It will "work to develop examinations that could be given at the end of the courses so that colleges, employers and students could judge how much learning had taken place. Course developers would be asked to consult with colleges on standards, so that the offerings could be created with the goal of having credit transferred to many institutions."

This is in line with the growing popularity of the Career Readiness Certificate, which "is mostly used for jobs that do not require a bachelor's degree, but it is becoming common in many fields in which people are educated at community colleges." (Inside Higher Ed, The New Exit Exam... for Jobs, July 9, 2008). According to the article, the CRC was first introduced in 2004 in Virginia and has now been recognized by 16 states; a community college that requires a certain level of the CRC (there are three levels, depending on the test-taker's demonstrated competency: gold, silver or bronze) for admission into its pharmacy technician program has seen the dropout rate go down from 51% (fifty-one per-cent!) to almost zero. Other community colleges "are considering testing all incoming freshman for the CRC, to see where they stand academically" and adapt to their level of preparedness.

The "US Push for Free Online Courses" article also gives details about the job training and the loan programs, which involve a lot more money than the project on online courses (the description starts at the headline "Support for Job Training and Facilities"). The job training program will give money directly to community colleges for the first five years (at the rate of $500 million a year), and then will give money to both states and community colleges ($1.3 billion a year). "To be eligible, community colleges would need to agree to track and report on student outcomes, and to set targets for graduation rates and 'employment-related outcomes,' while also serving 'high need populations.' " For the loan program, the government will give money to the states, which will then distribute funds based on the demonstrated needs of community colleges. While this is all done for a very worthy cause, the amount of bureaucracy it will involve is breathtaking.

Of course, not everyone thrives in a self-learning environment, and one can wonder whether online courses with no instructor are the tool high-school dropouts most need to improve their situation. On the same day it published "US Push for Free Online Courses", Inside Higher Ed just so happened to publish an article entitled "The Evidence on Online Education". The lead paragraph is: "Online learning has definite advantages over face-to-face instruction when it comes to teaching and learning, according to a new meta-analysis released Friday by the U.S. Department of Education."

But if you read the fourth paragraph (after three paragraphs on the positive aspects of online ed), you learn that "[a] meta-analysis is one that takes all of the existing studies and looks at them for patterns and conclusions that can be drawn from the accumulation of evidence." In other words, no new study was conducted. The analysis is on 51 previous studies, with no information on what the studies were about and what specific topics they considered. (For instance, 51 studies on first-year physics courses suggest different conclusions than 10 studies on humanities, 15 on science and 26 on engineering.)

There is more. "[T]he report attributes much of the success in learning online (blended or entirely) not to technology but to time." The author then quotes an excerpt of that report, including: "[O]ne should note that online learning is much more conducive to the expansion of learning time than is face-to-face instruction." In other words, online learning is not what is making a difference. Personalized learning is.

Mathematical Models in Finance and Economics

"In Modeling Risk, The Human Factor was Left Out." (New York Times, November 4, 2008) This article argues that, while the financial crisis has been blamed on the mathematical models of financial engineering, "the larger failure, [experts] say, was human - in how the risk models were applied, understood and managed." There is a good analysis of the role credit-default swaps played in the current situation, and a brief mention of the computer credit-scoring models that replaced human judgment in approving mortgages.

One of the people quoted also touches upon the difficulty in quantifying factors like "complexity, transparency, liquidity and leverage." In addition, a report by the Fed suggests Wall Street analysts "correctly predicted that a drop in real estate prices of 10 or 20 percent would imperil the market for subprime mortgage-backed securities", but severely underestimated the probability of this happening. (By now, errors in estimating probabilities are nothing new.) The article has many more nuggets of wisdom, which I will let you discover for yourself, but I'll end with this excellent quote by Columbia University Emanuel Derman: "To confuse the model with the world is to embrace a future disaster driven by the belief that humans obey mathematical rules."

"Bernanke's Models, and Their Limits." (New York Times, October 30, 2005) This article states that Bernanke, who was at the time the nominee to replace Alan Greenspan at the Fed, "has focused on the use of mathematical models to set monetary policy" and "has written repeatedly about ways of using mathematical models of a dauntingly complex economy to set monetary policy." His philosophy, according to the article, is "that the central bank should use a model, not just hunches, to decide about interest rates and the money supply."

A late economist at MIT, assessing one of Bernanke's models published in the literature, commented that "the interesting issue is not the gentle part of the trip but rather when it crashes;" the model apparently did not account for the fact that credit and liquidity could dry up. (And this was 2005.) But another professor, this time from Berkeley, supported Bernanke: "He of course understands that even in normal times, the best model is just a guide."

"The Anti-Macroeconomics Roar Grows Louder." (New York Times, June 3, 2009) Steven Levitt, of Freakonomics fame, argues that macroeconomics isn't about "writing down fancy mathematical models." Part of the problem, he says, is a lack of macro data, which makes it hard to measure the skills a good macroeconomist is supposed to have. He writes: "The single easiest way to make a mark in a modern macro paper is to solve a problem that is really, really hard mathematically. Even if it is not that relevant to anything, it is seen as a sign that the author has “impressive skills,” which is enough to get a job — and even tenure sometimes — at top universities."

While mathematical models are hopefully more relevant in real life than Levitt implies, he voices a valid concern, which is also illustrated in the chasm in my field between "hard operations research" (which is mathematics-driven) and "soft operations research" (which is non-mathematical). You can read more about that in OR/MS Today, especially "Taming Hard Problems with Soft O.R." (April 2009), "The Case for Soft O.R." (April 2009) and "Who's SORiors now?" (June 2009) The battle goes on.

Car windshields and Styrofoam cups

Diandra Leslie-Pelecky, guest blogger at Cocktail Party physics, has an interesting post, complete with equations and back-of-the-envelope estimates, on whether a Styrofoam cup can break a windshield. Her instinctive answer is: of course not. It's a Styrofoam cup! Then she saw the pictures - she was asked the question by a local reporter after someone in a car discarded a cup of soda through his or her window, cup that then smashed the windshield of a poor woman who happened to drive by in the opposite direction.

For those of you who can't get over the fact that all this damage was caused by a Styrofoam cup, the key isn't the cup (very light), but the soda that was in it (much heavier). As Leslie-Pelecky writes, "We know it was a cup of soda because, in addition to poor Marilyn [the innocent driver] getting wet, there is sticky dried cola-colored liquid all over the inside of the car.  The police pulled part of the cup out of the windshield (which is how we know it came from Sonic) and found a shredded straw inside the car." (The driver was lucky to be wearing glasses,which "were pitted from where pieces of windshield glass hit them.")

Leslie-Pelecky proceeds to estimate the force with which the cup full of soda hit the windshield, coming up with a range of 24-120 pounds because of the lack of data on the angle at which the cup hit the car and other parameters. It's a fun article to read for people who like science; Leslie-Pelecky points out that more kids might become interested in physics if high school exercises were more relevant to real life. But when it comes to throwing your cup of soda out of your moving car, the moral of the story is that it's a big no.

Manufacturing and America's Competitiveness

The July/August issue of Harvard Business Review has an excellent article on "Restoring American Competitiveness" by Gary Pisano and Willy Shih (pp.114-125). The authors object to the commonly held view that outsourcing low-tech manufacturing has had no impact on the American ability to compete; on the contrary, they argue that "decades of outsourcing manufacturing has left US industry without the means to invent the next generation of high-tech products that are key to rebuilding its economy."

The statistics they provide certainly give pause. For instance, the authors point out that the US trade balance in high-tech products has steadily declined, from a surplus in 1999 and 2000 to a deficit from 2002 onwards; the deficit now reaches 53.6 billion dollars. (See Figure 6-19 in the National Science Board report, "Science and Engineering Indicators 2008".) Furthermore, "average real weekly wages have essentially remained flat since 1980, meaning that the US economy has been unable to provide a rising standard of living for the majority of its people."

One of the authors' central claims is the importance to high-tech innovation of what they call "the industrial commons", i.e., "R&D know-how, advanced process development and engineering skills, and manufacturing competencies related to a specific technology." Such a commons is often rooted geographically; for instance, US biotech firms have clustered in Boston, San Diego and San Francisco, while Germany is known for its mechanical engineering commons due to its automobile and machine tool industries. This is why, in the authors' eyes, the world is not flat at all. (Sorry, Thomas Friedman.) Instead, "innovation in one business can spawn whole new industries," which makes it difficult, for companies that have outsourced the manufacturing of mature products, to compete in more cutting-edge areas based on similar technologies. (See the example of the solar panel industry p.118.)

An example p.119 also shows the erosion of the US industrial commons in the personal computer industry; I was surprised to learn that "[n]early every US brand of notebook computers, except Apple, is now designed in Asia, and the same is true for most cell phones and many other handheld electronic devices." (p.116 - italics mine)

A couple of additional quotes:

• "One [myth] is the popular belief that an advanced economy like the United States no longer needs to manufacture and can thrive exclusively as a hub for high-value-added design and innovation. In reality, there are relatively few high-tech industries where the manufacturing process is not a factor in developing new - especially, radically new - products." (p.119)
• "[The prevailing view that the migration of mature manufacturing industries away from developed countries like the United States is just part of a healthy, natural process of economic evolution] ignores the fact that new cutting-edge high-tech products often depend in some critical way on the commons of a mature industry" (p.120)

For other great insights, including (1) a sidebar on "why Amazon's Kindle 2 can't be made in the US", (2) a discussion on the role of government, especially in funding basic vs applied research ("Government funding for applied research has declined even more sharply [than for basic research]... [But] Creating the internet involved little or no new basic science. It did, however, require significant investments in applied research," p.122), and (3) words of advice to companies ("Create technology-savvy boards of directors," p.125), head for the closest newsstand while the HBR issue is in stores!

For the opposite view, check out "Innovation in America: A Gathering Storm" from the November 20, 2008 edition of the Economist, which refers heavily to Amar Bhide's "The Venturesome Economy," a 2008 Best Business Book of the Year. The book argues that the US doesn't really need more PhDs in science and engineering: innovation happening abroad will find its way to the States due to the appeal of the large American customer base (so companies will want to market their products in the US), and bringing high-tech innovation to consumers requires good business skills rather than advanced scientific skills. From the Economist: "For America to retain this sort of edge, then, what the country needs is better MBAs, not more PhDs."

Bhide "argues that the obsession with the number of doctorates and technical graduates is misplaced because the “high-level” inventions and ideas such boffins come up with travel easily across national borders." This assumes that all the inventions that aren't discovered in the US can be discovered elsewhere at about the same time. I disagree with the idea that geographic location doesn't matter. The Academic Rankings of World Universities are consistently dominated by American institutions; US-based professors have often spent decades developing cutting-edge expertise, which cannot easily be acquired in a short period of time by others. Of course, some non-US universities, such as the University of Utrecht in the Netherlands and Sapienza University of Rome in Italy, rightfully enjoy worldwide fame, but making it more difficult for US-based professors to supervise students due to the lack of funding impedes their ability to make discoveries and pass along their knowledge; it also hurts the students who are unable to realize their full potential. This, in turn, delays innovation. Hoping research is "background-independent" and will simply be performed in another country before finding its way to the States is naive, at least for now.

The comments at the bottom of the Economist article widely slam the position that the US doesn't need more PhDs. I wouldn't throw away the HBR article just yet.

First Book

Many charities have seen the amount of their donations decrease this year and, as the end of their fiscal year draws near, are making last-minute pleas for help before the June 30 deadline. So I thought I'd make a quick plug for my favorite charity, First Book, an organization that gives new books to children in need. Every ten dollars donated provides five new books to kids of low-income families; in addition, First Book has a four-star "exceptional" rating from Charity Navigator, which means the money donated really goes toward buying books for the kids, and not, say, toward paying the staff's salary. (First Book extensively relies on volunteers to fulfill its mission.)

Fighting illiteracy is a cause dear to my heart, because - as the people who have been reading this blog for a while know - education helped my family get out of poverty many years ago. My grandmother was a genocide survivor who arrived in France with only the clothes on her back; my grandfather (her future husband) was a blue-collar worker who painted cars before there were robots to do that, and before people knew they should wear masks to protect their lungs from the paint fumes. He died in the hospital from respiratory problems. There is a similar story on the other side of the family. I wish my grandmother, who outlived her husband by two decades, had lived long enough to learn that I had been admitted to MIT - she missed the big news by a matter of months. For her, the school cook who hadn't graduated from high school, that would have been like her granddaughter being elected President. Anyway, my parents got where they are by studying very hard in school, and making sure their children studied very hard too. If kids don't know how to read, though, they will never get a chance to improve their life. High levels of literacy help give disadvantaged children and their family the tools to take control of their situation and fight for themselves.

So I donate to First Book every year, and I hope you'll make a donation too. The minimum amount is just five dollars.