Previous month:
July 2008
Next month:
September 2008

August 2008

College Board and Project MINDSET

Following my last post, where I wondered what was "wrong with taking college-level courses in college and not before", someone sent me (thank you!) a link to this radio report on, hang on to your seats, the College Board's decision to "roll out a new college assessment exam for eighth graders in 2010." The Los Angeles Times wrote an excellent article on the topic, and includes the quote, by someone from the Princeton Review: "Eighth grade is not the key year for college assessment. That's sixth grade." Why not kindergarten while you're at it? Although the College Board pretends that "This test will help schools identify students who have some talent and could likely succeed if they take honors or AP courses, but have not been recognized," I find myself siding with the critics, who believe this "would just boost the pressures for students considering college" and is a way for the College Board to increase its revenue by taking advantage of the country's obsession with testing.

I'd much rather see some of this energy spent on improving high school education, and that is exactly the goal of three universities leading the NSF-sponsored Project MINDSET (Mathematics Instruction using Decision Science and Engineering Tools). The idea is to design a math course for high school students that is "fully applications-based and problem-driven". Apparently, there is a severe need for that, or - more generally - anything that will help the students do better in math. The authors of the article in the August edition of OR/MS Today, linked above, mention the following question at the third NAEP mathematics test: "An army bus holds 36 soldiers. If 1,128 soldiers are being bussed to their training site, how many buses are needed?" According to the article, only twenty-three percent of the students gave the correct response. (In case you're wondering, the correct answer is 1128/36 rounded up, or 32.)

I felt the authors were doing themselves a disservice by starting the article with grand sentences such as: "Can you imagine high school students challenging their local police chief on the merits of the Poisson process and its impact on staffing 911 and police patrol units?" and "Can you imagine high school women excited about using the principles of queueing theory to petition their state legislature to adopt the public building codes of New Zealand that were designed to reduce waiting times for bathroom facilities at concerts and theater events?" but since the article is intended for the operations research and management science population, I suspect they were trying to address the geeks in us (operations researchers). My main complaint about the report is that it gave no example of this new textbook's contents - it would have been interesting to compare two approaches (the traditional one and the Project MINDSET's one) for the same educational goal.

Much better in my opinion for a general audience (and again, the audience for OR/MS Today isn't exactly "general") is the press release written by Wayne State University to describe the project. It explains that applications of this new "12th-year mathematics curriculum" include "designing school bus routes, finding the best location for a new recreation center, calculating an appropriate automobile insurance policy deductible or evaluating ways to reduce wait lines for public restrooms." In that respect, it reminds me of some of the industrial engineering projects my department runs for the Introduction to Engineering course, which provides first-year students with a hands-on overview of the various majors. It makes sense that, if high school teachers receive the proper training, some of the projects could be completed by seniors. That would fundamentally change the way students think about operations research.

Unfortunately, much remains to be done for this new course to become widely adopted, in particular at the school board level. In the meantime, as another potential funded project, I'd love to see someone create a free web-based assessment exam for students finishing middle school. After all, you don't need a lot of different tests if there is no incentive for cheating and the grade doesn't count for scholarship applications. Maybe if companies like the College Board lose their ability to profit from worried eighth-graders, they will leave those kids and their teachers alone.

Education in Pennsylvania

The "Tapping America's Potential" website has a very well-done map showing US students' performance state by state. If you put your mouse over Pennsylvania - my current home state - you learn for instance that "only 31% of [PA] 8th graders were at or above proficiency in math in 2005." That doesn't exactly inspire confidence. (Even worse, the percentages for other states aren't that different, and the national average is 29%.) Clicking on Pennsylvania opens this pdf document, which contains fascinating statistics about Pennsylvania and the country as a whole. For instance, only about half middle-level science teachers and math teachers are certified, and only about half high school chemistry teachers have their certification in chemistry. (Incidentally, my first-year advisees, who all come from different hometowns in various parts of the country, all agree that they hated chemistry in high school because their teacher was incompetent, but don't have nearly as severe a reaction toward math or physics.) Three quarters of high school math teachers have their main certification in math, which raises questions regarding the quarter who doesn't, but at least the situation isn't as dire. Many engineering challenges, such as providing access to clean water or engineering better medicine, require a knowledge of chemistry; it is a pity the topic is emerging as the most disliked subject in high school.

I was intrigued by a number of other stats in the pdf file - only 7% of high school students in the country (6% in PA) take the Advanced Placement test in calculus, for instance. That is one of those numbers for which it is difficult to decide on a "right" value, though. When I read the number in the middle of all the other stats (SAT scores and the like), which should plainly be as high as possible, my first reaction was that this percentage should be higher too. Higher is better, right? But that would assume it is a good thing for high school students to take college courses early. The Collegeboard website pompously announces: "AP can change your life. Through college-level AP courses, you enter a universe of knowledge that might otherwise remain unexplored in high school; through AP Exams, you have the opportunity to earn credit or advanced standing at most of the nation's colleges and universities." 

My guess is that a lot of students in AP courses are college-bound, eager to take a head start. Is that a good thing? What's wrong with taking college courses in college and not before? Having a mediocre or average high school teacher for the AP calculus course might result in a student passing the test but struggling once he attends a university. A 2006 survey by Harvard and the University of Virginia even established that "high school AP courses do not predict college success in science." In the words of one of the study's authors, "it appears that the educational benefits of an AP science course as opposed to a regular high school honors course are smaller than students and teachers have been led to believe." The number of AP tests has tripled since 1995, though, maybe because parents feel pressured to help their children keep up with the competition. But it is always easier to learn something correctly the first time around, than to learn it imperfectly and try to correct bad reasoning habits later on. In that sense, AP courses can do more harm than good.

As a side note, my readers living in Pennsylvania might appreciate a website I was recently told about - salaries of teachers in public schools are a matter of public record under Pennsylvania law, and it did not take long for an organization annoyed with teachers' threats of strike to put all the salaries on the Internet, with the teachers' names and the highest degree they have received. The database is apparently maintained by the NJ-based Ashbury Park Press newspaper, and might be of interest for those of you who have children educated in PA. For instance, the salaries of math teachers in the Bethlehem Area School District are here. Click on "New Search" at the bottom to run your search. Enjoy!

Lofty Goals

I recently came across the "Tapping America's Potential" website, whose self-proclaimed goal is to "double the number of science, technology, engineering and mathematics graduates with bachelor's degrees by 2015." In the progress report, the prominent businesspeople behind the site quickly acknowledge that the 2015 number of graduates will fall abysmally short of target - they phrase it in a more measured way ("U.S. STEM Bachelor's Degree Production Not On Track to Meet TAP Goal", p.5), but the graph on the same page speaks volumes. Over the past five years, the number of STEM graduates who are US citizens or permanent residents has remained between 220,000 and 225,000 every year, while the target is to reach 400,000. The report's authors give convincing reasons for this ambitious goal:

  • "The Bureau of Labor Statistics projects that employment in science and engineering occupations will grow 70 percent faster than the overall growth for all occupations." (p.6)
  • "In 2006, the already low unemployment rate for STEM graduates dropped to 2.5 percent, and starting salaries were higher for students graduating with STEM degrees, particularly those with engineering degrees, than for most other majors." (p.6)
  • "There is a desperate need for STEM majors to teach math and science at US schools. Research indicates that a highly qualified teacher is one of the most important factors in raising student achievement."

The report mentions the 2007 America Creating Opportunities to Meaningfully Promote Excellence in Technology, Education, and Science Act (yes, that makes America COMPETES), which - among other things - attempts to improve education in math and science by enabling part-time master's programs for current teachers to "improve their content knowledge and pedagogical skills" (Title VI). The tap2015 progress report highlights that this act increases support for the Robert Noyce scholarship program, "which provides scholarships for STEM undergraduate majors who agree to become K-12 math and science teachers" (interestingly, the scholarship is for institutions to provide funds for students, rather than for students themselves - adding a layer of bureaucracy is never a good thing, in my opinion), as well as "increasing the number of teachers prepared to teach Advanced Placement and pre-International Baccalaureate college preparatory courses, and providing Math Now grants to improve elementary and secondary math instruction." Sadly, in the end, despite all the consensus and bipartisan agreement, "appropriations provided either flat funding or real declines in fiscal year 2008".

Page 11 of the report provides the country-by-country scores of the mathematics literacy scale and combined science literacy scale as determined by the Organization for Economic Co-operation and Development (OECD). The United States perform below the OECD average and (for both scales) below countries such as Finland, Canada, Japan, Ireland, Austria, Australia, Belgium, Switzerland, Sweden, Germany, the Czech Republic, the Republic of Korea, Denmark, the United Kingdom, Poland and Hungary. And yet, few of these countries are regarded as cutting-edge technological innovators; their best students come to the United States to get a Master's or a doctorate, and oftentimes do not return to their home country. K-12 education in the United States is notoriously lacking, and has been lacking for decades. Maybe it is unrealistic to expect a lot of instructors who have been educated in the American system - especially those in the second half of their career - to teach math and science in middle and high school at a level that would bring the country to par with its competitors. After all, the veteran teachers educated the students now deemed to perform below OECD standards. Also keep in mind that students from small countries, such as Denmark or Finland - in addition to ranking above the States on the math and science rankings - receive top-notch training in English; the justification is that their own country is so small its residents are bound to interact with many foreigners. In the same way that the US are now importing a large fraction of its doctoral students in science and engineering, it might start to consider importing math and science schoolteachers.

Log-Robust Portfolio Management

I wrote a while back about my work on log-robust portfolio management with my doctoral student Ban Kawas. We have just posted the revised version of the paper on Optimization Online - check it out! We are very excited about the changes we have made to the paper: we have derived new theoretical results, added more insights, and strengthened the literature review. I describe the changes in more details below.

First, let me briefly go over the ideas behind log-robust portfolio management. There is a large body of literature explaining the limitations of the Lognormal model of asset prices and debating the correct choice of the underlying distribution. (The definition of the Lognormal distribution is here, and its implication for the Black-Scholes option pricing model here.) An issue is that the Log-normal distribution underestimates the probability of tail events. In other words, the real-life, unknown distribution has fatter tails than the Log-normal model suggests, which leads to under-estimating downside risk. The debate on the correct distribution is still active, and will probably be never settled; however, one thing financial engineering researchers agree on is that randomness affects stock prices at the level of the continuously compounded rates of return (the logarithm of the ratio between the current stock price and the one of the previous time period). This means that, when you consider the continuously compounded rates of return, you can isolate random variables that depend neither on the length of the time period considered nor the stock prices' parameters.

Taking into account that portfolio managers are risk-averse and want to protect wealth against extreme events, this makes the setting very well-suited for a robust optimization approach. In robust optimization, random variables are modeled as uncertain parameters belonging to a known interval, and the scaled deviations of these parameters from their nominal values are constrained, for instance through the use of a budget of uncertainty. It is important to use independent uncertainty drivers, because the whole philosophy behind the approach is that not too many of these uncertain parameters can take their worst-case values at the same time. Robust optimization can be viewed as "worst-case optimization over a reasonable set of uncertain parameters;" the 'reasonable' part of this statement is what makes the approach avoid over-conservatism. The validity of this framework is supported by many numerical studies. Famous early papers on robust optimization include this one by my former PhD adviser and one of his then-students, and that one with a focus on ellipsoidal uncertainty sets.

One of the reasons robust optimization has not been widely extended beyond linear models, where the problem is linear in both the decision variables and the uncertainty, is that the tractability of the approach hinges on rewriting a max-min problem as one big maximization problem, using strong duality to reformulate the inner minimization problem over the uncertainty set as a maximization one. To keep the approach tractable in nonlinear cases and allow for the use of strong duality, one needs an objective function that is convex in the uncertainty (because it is going to be minimized) and concave in the decision variables (because that part is going to be maximized). If you don't remember what the big deal about convexity/concavity is, it guarantees that the algorithm stops at a solution that is globally optimal and not just locally optimal. Needless to say, there are not a whole lot of nonlinear problems out there with that specific structure, so robust optimization researchers have tended to focus on multi-stage linear programming problems instead, with concepts such as adjustable linear programs and finite adaptability. And then there is Ban and me, and our portfolio management problem that not only is important for practitioners, but allows for theoretical results, due to its convexity in the uncertainty parameters and its linearity (hence concavity) in the decision variables. In particular, our work leads to linear programming formulations, which can be solved efficiently, including for large-scale settings. This makes the approach very appealing in finance, where a manager can choose from hundreds or even thousands of stocks.

Besides the literature review, we have made two important additions to the paper: (i) we have strengthened the theoretical results, for instance showing that, in the case where stock prices are uncorrelated, the amount of money invested in each stock is inversely proportional to the standard deviation of the continuously compounded rate of return, and the optimal scaled deviations do not reach their bounds, which streamlines some of the proofs, (ii) we have added new insights, in particular regarding why diversification occurs (that is, why the number of stocks the manager invests in increases as the budget of uncertainty increases up to a certain point, and then starts decreasing). We are grateful to an anonymous reviewer for asking such good questions about the work.

We are very proud of this paper, which we feel addresses an important drawback of current theoretical models and should help portfolio managers make better decisions. We also have extensions about short sales and derivatives, which we will be posting soon. This work is currently funded in part by the National Science Foundation Grant CMMI-0757983.

Business Models, Jim Collins, and Michael Porter

I've enjoyed reading this article of The Economist dated August 14, 2008, entitled "Bank Strategies - No Size Fits All." Its main argument is that bank management (as in: people) bears a lot more of the responsibility for the current crisis than business models. The article's focus isn't on quantitative risk models, but rather on whether investment banking, wealth management and asset management should be separated. "The structure of an organisation matters less than the quality of the people who lead it. For bank regulators and shareholders, the question is less “what?”, more “who?”." One could draw the same conclusions when it comes to computer models: they are only as good as the people who feed them data and analyze them.

Also, Jim Collins is profiled in the August 8th installment of the series on management thinkers. We owe him "Good to Great" and "Built to Last", two of the best-selling business books of all time. (If only everybody's research could be that influential.) The previous week saw Michael Porter being recognized for his contributions to strategic management, and in particular his framework of "five forces", now taught in business schools everywhere. I read his book "Competitive Strategy" a while back and I agree with the Economist reporter that the work is "academic to a fault," which is a polite way to say: not exactly geared toward the general public. You won't find any anecdotes here; the frameworks are abstract and described in such a dry prose - in Arial font, no less; when is the last time you saw an Arial font used for something else than a term paper? - that you feel immense pride if you read the book to the end and don't give up halfway. (Competitive advantage also happens to be the focus of August 4th's installment in The Economist's series on management ideas; you can read more about the "five forces" here.) In fact, Porter might not care about bestseller-dom, since he is already extremely famous in top management circles. The same Economist magazine writes, in its issue dated August 7th, 2008 about a book about Harvard Business School: "[The book's author] is a fan of Michael Porter, the nearest thing management theory has to a rock star." That makes you wonder about the super-stardom Porter might have achieved if he had bothered to make his books a bit more accessible to the rest of the world. 

Ten Days Left

It's hard to believe the Fall semester starts in ten days. Where did the summer go? Last week I attended the training for the First-Year EvoLUtion seminar series (which I finally decided to volunteer for), and next week I have the training for the book discussion, a meeting for all the first-year advisers, and then of course the meeting with my first-year advisees this coming Friday - Thursday being the big day where freshmen and freshwomen arrive on campus. My list of advisees is even on Banner already! (Banner being Lehigh's web system where faculty enter grades, look up class rosters, etc.) Assignments used to be made depending on the first initial of the last name, but the process seems to have changed - I have one advisee whose last name starts with a B, one with a D, one with a G, two with a K, three with a M, one with a N, two with a S, and two with a V. If there is a rule behind this, I haven't found it.

Thirteen is a good number - not too many so that I can keep an eye on them, not too few because there are 1,208 incoming students (not all of them in the College of Engineering), and less students for me means more students for someone else - the Registrar's Office typically assigns ten to fifteen students per adviser. This year I think I'll have all of my advisees fill a little form about themselves while we go to lunch. Last year students were told to go and meet with their adviser if they had an issue with their courses, but since most of them didn't have any issues, there is a whole bunch I barely spoke to when we were waiting in line for lunch in the dining hall and didn't see again until November when we met to discuss Spring registration (because at lunch no table was big enough to seat all of us and they went to sit in little groups). That is obviously not optimal, especially for the students who end up struggling with their classes. For instance, quite a few first-year students can tell when they're headed for academic trouble (not because they're troublemakers, but because calculus is way over their head), but they're reluctant to sign up for tutoring unless their adviser (as in: me) agrees with them, prints the form available on the webpage on her office printer, makes them fill it in front of her, locates the office where they're supposed to bring the form on the Lehigh website, and sends them on their way. Total time for me: two minutes and a half. Impact for the students of not having to repeat their first math course in college: priceless. But they're never going to show up in my office to talk about their math troubles if they haven't met me already. So this year they'll have to stick around during lunch even if they don't want to! I also might start weekly office hours for my first-year advisees, so that they know a time of the week where I'll always be there and have time for them if they want to drop by.

There was such a high level of energy in the room for the EvoLUtion training - it is amazing to see how many faculty and staff members volunteer for those things when they have nothing to gain in return, except the satisfaction of helping college students - or at least trying to. The staff members in the Office of the First-Year Experience in particular are the enthusiastic souls who designed the whole program, which only started at Lehigh four years ago (and in the first year there was no evoLUtion seminar and the book discussion was optional.) So it is great to talk with the people who got this off the ground. We discussed characteristics (such as "hardworking" or "nice") vs sources of identity (like religious or spiritual affiliation, dietary needs, artistic skill, or athletic status), the Inputs- Environments- Outputs theory and all kinds of exercises we'll have the students do. I selfishly enjoy that kind of training a lot, not only because I'm interested in psychology (which I am), but because it provides lots of opportunities to interact with staff members I'd never meet otherwise. For instance, I ended up being seated next to someone in charge of the co-op program, who asked me why my department (Industrial and Systems Engineering) was sending so few students do a co-op this year, while companies absolutely love our students (the quantitative but business-oriented training of industrial engineers is in high demand). The transitions in the department might have had something to do with it, but they are a thing of the past now, and it's time to revive the co-op program.

So I am teaching two courses this coming semester - I am very much looking forward to that! I've missed teaching. (Last semester I didn't teach but only supervised theses and projects.) I've taught IE 316 every Fall semester since I joined Lehigh; it's a senior-level elective about optimization models in real life. There are 45 students registered so far, and I am so grateful that enrollment is not exceeding 47 because that's the maximum capacity of the biggest classroom in the industrial engineering building - having more than 47 students enrolled would get us kicked out and would send us trekking to some dreary old place like the mathematics building, Christmas-Saucon. (Sorry math folks - I taught in Christmas-Saucon my first semester at Lehigh and don't want to do it again.) Right there, the course is off to a good start.

I plan to cover problems in inventory management, revenue management, portfolio management, and the like; the syllabus has plenty of information on that. As I mentioned in an old post of mine, I will have more but shorter assignments this year, and I will also give the solutions to last year's exercises in the course packet for the students who want more training. The one thing I have to do before the semester starts is to re-do all the examples using GAMS rather than AMPL - students who took the prerequisite course in the Fall learned AMPL, which is the software I know, but those who took it in the Spring learned GAMS, which I know nothing about. So to give students the option to use the software they've learned before, I have one week to teach myself GAMS. I doubt it's very different from AMPL, though. I was counting the number of lectures we'll have and trying to see how many assignments I could squeeze into the semester, when it occurred to me that Thanksgiving is falling very late this year, so the week after Thanksgiving is also the last week of classes. Usually, there is one week between Thanksgiving and the last week of classes where I can schedule a quiz, but you can't schedule exams during the last week of classes, which means the last day I can have an exam is the Tuesday before Thanksgiving, and that is simply not going to happen. (Thanksgiving break starts on Wednesday here, which means, for some reason I've never understood, that students start leaving on the Tuesday before Thanksgiving, if not before.)

No big deal, I thought, I'll just schedule it for the Thursday before Thanksgiving week. I went back to planning the homework schedule, but I had this nagging feeling something was really, really off. Then I remembered: the week before Thanksgiving is Lehigh-Lafayette week! (a week of parties for the students before the big football game on Saturday.) Gasp. Scheduling an exam for that Thursday is just cruel. What if the Lehigh team actually wins (they haven't won since before I joined, which as far as I'm concerned means that they haven't won since the beginning of civilization) and the 45 industrial engineering seniors taking my course remember for the rest of their life that I prevented them from enjoying the celebrations leading to that game, ruined their senior year, ruined their college memories, and then sent them into a downward spiral of abysmal depression? More seriously, if I schedule a test for that Thursday, I might be the one sent into a downward spiral of abysmal depression. After all, people do a lot better on exams when they can keep both eyes open at the same time. Anyway, I'm planning to have the test on the Tuesday of the Lehigh-Lafayette week, and since we can't just be done with the course mid-November, we will also have a project about how companies really do implement optimization models in real life. All this to explain to the students who are reading this blog why the last test will be so early this year and why we'll be having a final project. I know the rest of the world is just fascinated.

The other course I will be teaching is IE 406, a course on linear optimization for first-year PhD students, which I've never taught before (one of my colleagues teaches it every year, but he'll be on sabbatical this academic year), but I took the course from the co-author of the textbook Lehigh uses, my very own former adviser at MIT, Dimitris Bertsimas. Besides, linear optimization is a central component of my research, so if there is one topic I know by heart, it is this one. I even have all the lectures and assignments preciously saved from the Fall of 2000 when I took the course as a student at MIT, which is nothing short of a miracle, given my ability to mix papers and forget where I've placed them. (And the fact that I only have one paper copy and no electronic copy. Once I located the lecture notes over the summer, I put them all in one big plastic bag, which has become one of my most prized possessions. If I could, I'd put a RFID chip on it.) So this semester should be very enjoyable, I think. I'm really, really excited about teaching this semester - it's going to be a great term!

Left Behind in the Valley

According to my local newspaper in its August 14, 2008 edition, "Fewer schools in the Lehigh Valley and across Pennsylvania are hitting federal academic goals as target scores for math and reading scores have risen, according to a state academic achievement report released today." The requirements of the No Child Left Behind Act have never struck anyone as particularly stringent, and yet students keep underperforming. The decrease in this year's numbers was attributed to the federal government raising the bar: "This year, the state required 63 percent of all students attain proficient scores or better in reading, up from 54 percent. In math, 56 percent of the students had to be proficient or better in math, up from 45 percent." You mean, success is achieved when a third of Pennsylvania students can't understand what they read? No wonder the state is in bad shape. People often complain that test-based accountability makes teachers "teach to the test" instead of helping students truly learn the material, but if that's the kind of results teaching to the test gets, we've got a while to go before we worry students aren't learning the right type of skills. Right now, many aren't learning a thing, period.

I became curious to see the actual test results, and the Web being the wonderful resource that it is, it won't come as a surprise that the school-by-school, detailed test scores are indeed posted on the Internet. Two websites provide this information: one from the Pennsylvania Department of Education, and one with data for all the states, which will be of interest also for the non-Pennsylvanian readers of this blog. (The following links point to the second, country-wide website.) The disparities in the scores for the Bethlehem Area School District, for instance, are quite simply staggering. Take Broughal Middle School, which I can literally see from my office window, on the opposite corner of the intersection between Broadhead Ave and Packer Ave. My jaw dropped when I read that 83.5% of its students are considered economically disadvantaged. 83.5%! Now, I knew there were a lot of students from low-income families, because last year Lehigh University organized a coat drive to clothe the Broughal kids during the winter months, but I'd never have thought that the percentage of disadvantaged kids was so high. And then you wonder why their parents don't like us Lehigh people.

You can read more statistics on Broughal here, including the fact that over two-thirds of the kids at Broughal are Hispanic, which, in the Bethlehem's South Side, happens to be a good indicator of economical disadvantage (also click on the "Community Demographics" tab - the amount of data gathered on each school district is extensive: for instance, the median age in the Bethlehem Area school district is 38, 14.2% of the households have an income of $15,000 or lower, and 31.7% of adults have at least a Bachelor's degree.) Anyway, 45.4% of the Broughal kids displayed reading proficiency and 52.2% displayed math proficiency. The other two schools on the South Side (both elementary schools with at least 70% of economically disadvantaged children) got low scores as well. While many Hispanics have successful and educated kids, the issue of integrating the not-so-successful ones will take on an increased urgency over the next few years, as the number of Hispanic students has been rising steadily and will keep on doing so. Helping minority kids is not just "charity work": as minorities become a bigger part of the US population, their level of education will have a bigger impact on measures of overall performance; improving the perception of the US educational system abroad requires spending resources on the kids who are struggling and who will, in the not so distant future, represent the majority of the American workforce.

According to this New York Times article dated June 1, 2007, "Hispanic students accounted for just 6 percent of public school enrollment in 1972, but by 2005 their numbers had grown to 20 percent." It turns out that this is a problem because of the language difficulties so many kids face when they grow up in a non-English-speaking household. Spanish is a beautiful language and Pablo Neruda might well be the best poet in the world, but the tests in American schools are given in English, and it is unfortunate that many kids don't understand the questions. On December 16, 2007, the New York Times ran a story entitled "No Child Left Behind? Say it in Spanish," which describes the predicament of suburban schools faced with an influx of Spanish students, and specifically of the Fernbrook school in Randolph, N.J. "The school [had...] passing rates in the 60s and 70s instead of the 90-plus numbers posted by the district’s other schools. Fernbrook seemed to be struggling to reach many new children of immigrants, especially Hispanics. [...] The number of Hispanic students at Fernbrook had doubled in five years, and there was concern that lower achievement had become chronic." (Residents worried that poorly performing schools would decrease their house values.)

The school district then decided to make a big investment in Fernbrook despite a tight financial situation, hired more teachers, "including one who spoke Spanish," as well as "a literacy coach to show [teachers] some new strategies to help students who were behind in reading." Also, "class size in kindergarten and first and second grades was reduced," "a full-time social worker was added, and the school’s reading specialist was made a full-time position." It was a zero-sum game: the improvements Fernbrook got, other schools in the district were deprived of. But it makes sense to allocate more resources to the school that need them most. The principal explains: "These kids come in with half the vocabulary of middle-class kids. We have six different categories of kids in terms of vocabulary, and they each get different instruction." For instance, "The group with the lowest skills gets a half-hour every day with the reading specialist." Of course you could argue that it would be a lot better if everyone arrived at school knowing English. It is not happening, though - there is no point in wishing the problem will just go away. In fifteen years, these kids will represent a significant part of the US workforce, and we need them to perform.

So what about Broughal Middle School? It is in an blighted area with lots of cheap housing. People who care about their house's value live on the other side of the river (Bethlehem's North Side), or the other side of the hill (Saucon Valley School District). A new school is being built right next to the old one - old, run-down Broughal is slated for demolition in 2009 - but the school district is struggling with a budget shortfall, supposedly because of the increase in enrollment (currently 15,000 students) and the increase in fuel costs. Board members have asked the contractor to revise his plans in order to cut construction costs. The odds of Broughal students don't look good.

Further reading:

  • "How well are they really doing?" New York Times, August 11, 2008. "The states have made a mockery of that provision [of documenting student progress in exchange for financial aid], using weak tests, setting passing scores low or rewriting tests from year to year, making it impossible to compare progress — or its absence — over time." and "Most state tests [in math] had no questions at the third for fourth levels at all. Worse, the teachers may be dropping more challenging math course work to focus on the low-level material covered in state tests."
  • "A high school finds itself left behind and drowning" New York Times, June 23, 2008. The "dismaying film [on Douglass High School] isn’t really asking whether No Child Left Behind can help Douglass. It’s asking whether anything can. The film [...] is filled largely with teenagers who are drowning in apathy and attitude."
  • "Report see costs in some academic gains" New York Times, June 18, 2008. A new study argues that "steady academic gains for low-achieving students in recent years [have come] at the expense of top students." An example: "In tests of fourth-grade reading from 2000 to 2007, the scores of the lowest-achieving students increased by 16 points on a 280-point scale, compared with a gain of three points for top-achieving students." I'll venture that maybe the two scores are growing at different rates because it is a lot harder to improve when you're already very good - the grading scale might not have room for a 16 point increase on that end of the spectrum. Just a thought.
  • "Math scores rise, but reading is mixed" New York Times, September 26, 2007. "The average math score for fourth graders [in the national reading and math assessment, which is more reliable than state tests] is at its highest level in 17 years, and the percentage of fourth graders scoring at or above proficiency rose to 39 percent this year, up eight points since the federal law took effect." 39 percent! Are we supposed to cheer?

Science: The New Cash Cow

In her latest column, a journalist at the Philadelphia Inquirer takes issues with the Franklin Institute Science Museum - the museum recently renamed itself "the Franklin" and has been organizing blockbuster shows such as "Pirates" that please the general public but tell kids little about science itself, and - even more annoyingly - has been charging patrons admission rates unheard of even in New York City. (Thanks to John Hunter who wrote first about that column in his blog.) The journalist explains: "Adult admission has soared to $23.25 if you want to see the current exhibits [...]. 'Adults' happens to mean anyone over 11, a rather severe view for an organization geared toward students. Imax? That's $5.50 extra. Audio tour? Yo ho ho, and an additional $6." In contrast, the American Museum of Natural History "offers $17 student tickets for ages 13 through 17 that include general admission and its more substantive featured exhibit."

People, I believe, are slowly waking up to the need for their children to be scientifically literate if they want to succeed in tomorrow's global economy - newspapers have been covering the issue of science in American education quite aggressively, with reason: there will be a dire shortage for scientists and engineers when the baby-boomers start to collect their pension, and American competitiveness hinges on the country's ability to replace the retirees. The New York Times, among others, has pushed that topic at the forefront of countless articles. For instance, Michael Janofsky writes in "Report Says States Aim Low in Science Classes" (December 8, 2005) that "The report also appears to support concerns raised by a growing number of university officials and corporate executives, who say that the failure to produce students well-prepared in science is undermining the country's production of scientists and engineers and putting the nation's economic future in jeopardy" and "Many states are not yet serious about teaching science."

In "Why American College Students Hate Science" (May 25, 2006) by Brent Staples, also of the New York Times, we learn that "Science education in this country faces two serious problems. The first is that too few Americans perform at the highest level in science, compared with our competitors abroad. The second problem is that large numbers of aspiring science majors, perhaps as many as half, are turned off by unimaginative teaching and migrate to other disciplines before graduating." The University of Maryland- Baltimore County, less famous than its College Park counterpart but just as determined, has put in place a merit-based scholarship program, called the Meyerhoff Scholars Program, where "the students are encouraged to study in groups and taught to solve complex problems collectively, as teams of scientists do. Most important, they are quickly exposed to cutting-edge science in laboratory settings, which demystifies the profession and gives them early access to work that often leads to early publication in scientific journals." The focus is on making sure the students learn the skills they need to excel: "Those who earn C's, for example, are encouraged to repeat those courses so they can master basic concepts before moving on." The article even states that "students who completed the Meyerhoff program are 5.3 times as likely to enroll in graduate study as the students who said no and went elsewhere."

While well-run programs such as the Meyerhoff one can attract students to scientific careers, there is no denying that academics have been ringing the alarm for years now - while researching articles for this post, I came across "National Panel Urges Earlier, Better Science Education" by one William. Here is an excerpt: "In far-reaching recommendations intended to improve science literacy, a national panel says students should be exposed to science at earlier ages, teachers should receive better training and curriculums should offer more real-life applications.The recommendations, released on Thursday, acknowledge that science instruction generally fails to give students challenging, inventive assignments and that students are often asked to memorize scientific theories and facts in a vacuum." And, later in the article: "The problem of uneven science teaching is particularly acute in the early grades, for which teachers are trained as generalists, educators said. In many elementary schools, educators said, science instruction tends to be watered down and given a secondary status in the school day." That does sound like an accurate description of the plight in today's schools, doesn't it? And then I looked at the date the article was published: December 7, 1994.

Another article I found states: "The National Science Foundation today opened a $25 million campaign to improve science education in elementary and secondary schools in the agency's largest such effort since the 1960's." A foundation official explains: "We are failing to provide an adequate background, an adequate introduction and an adequate level of science literacy for the population as a whole. As a result, the preparation of a sufficient number of candidates for careers in science, engineering and technology is inadequate." And later: "[The foundation official] said the project would at first focus on elementary school students because most of them received little science education, if any. [...] He said a major reason for the project was that the nation needed a ''technically sophisticated work force'' to remain economically competitive." The date of this article, entitled "Foundation Opens $25 Million Drive on Science Education"? January 20, 1987. At that speed, the country is in no danger of getting a speeding ticket.

Fortunately, the Internet has made it easier for concerned parents to gather data and research the issue - until recently, an article stopped to exist, so to speak, once the reader threw the edition of the day into the wastebasket. This means that the sorry state of science education isn't pushed as easily onto the sidelines. Moreover, the increased competition from other countries has added a sense of urgency to the situation, which might translate into some overdue improvements at long last. But what are parents supposed to do in the meantime? Those who did well in math and science can help their children with their homework. The others are faced with mounting pressure to provide some support for their kids, despite their own lack of scientific knowledge. Science museums appeal to those because they provide a structured environment to expose kids to science, and represent a fun family outing. (The kids learn, the parents learn - what is there not to like?) They will undoubtedly see their attendance rise over the coming years, as every family tries to do their part to help their children succeed. But this opens many business opportunities for the science-savvy, not all of which will have kids' best interests at heart. In the "good" category, John Hunter points out that the New York Academy of Sciences has a weekly podcast on "the ideas shaping science"; the Teaching Company, which has long focused on lifelong learning courses for adults, has started offering algebra, basic math, chemistry and geometry at the high school level. (The quality of their DVDs on art and music is simply extraordinary, so I am sure these new courses are outstanding as well. Just make sure you wait until they come on sale - a few times a year - before you buy them. The company's whole business is based on customers ordering during the sales, to allow for bulk production of each DVD. The topics on sale rotate throughout the year, so that there are a lot of courses offered at a discount at any given moment.) 

But others will certainly attempt to take advantage of many parents' feelings of inadequacy when it comes to math and science. "The Franklin" seems to be headed in that direction: let's milk the cash cow. Maybe we, as science and engineering educators, play a part in all this science-mania - we repeat that science education in the States is substandard, but schools are slow in showing progress; this naturally feeds parents' insecurity. In the end, I won't be surprised if throngs of well-meaning students find their way into the open arms of a parallel industry - similar to the SAT preparation business - where children with deep-pocketed parents have a chance to figure out at last what science is all about, while the rest of the kids must content themselves with teachers of varying quality. Of course, science in the States hasn't yet reached the level of popularity that would make that scenario come to pass any time soon. But in a couple of years, when baby-boomers begin to retire, the general public will take notice of the job openings and the salaries scientists and engineers will command (companies are always willing to pay a lot of money for scarce resources). Parents in turn will be a lot more willing to spend money on Junior's science education. If the school system hasn't started cleaning up its act, we will see the beginning of a very lucrative industry. The Franklin really is just the tip of the iceberg.

Risk Management on Wall Street

The August 7th issue of The Economist has an article about the crisis by an anonymous risk manager, who describes how banks became "so overexposed in the run-up to the credit crunch." Interestingly, the credit crisis is not the first time the market behaved in an unexpected manner in recent years - the author refers to the "hiccup in the structured-credit market in 2005," due to the downgrade of General Motors bonds, where, against everyone's predictions, "AAA tranches went down in price and non-investment-grade tranches went up." Sadly, while people were at a loss to explain that behavior, they didn't develop a mistrust strong enough of their valuation models to prevent the 2007 crisis, maybe because it was limited to one company.

The risk manager also explains that asset-backed securities such as CDO were viewed as credit-risk instruments by the market-risk department, while "the credit-risk department thought of them as market risk as they sat in the trading book." My goodness, these people sound like whiny little children. It seems incredible that no higher-up forced either department to take ownership of CDOs, but that is apparently what happened.

Later, the author lists three erroneous assumptions he made with his team:

  1. "all mark-to-market positions in the trading book would receive immediate attention when losses occurred, because their profits and losses were published daily."
  2. "if the market ran into difficulties, we could easily adjust and liquidate our positions, especially on securities rated AAA and AA. Our focus was always on the non-investment-grade part of the portfolio, especially the emerging-markets paper."
  3. "the rating agencies simply knew best."

He also mentions the immense "pressure on the risk department to keep up and approve transactions," and the unhelpful attitude of traders, who saw the risk management team "as obstructive and a hindrance to their ability to earn higher bonuses." His later comment on the risk managers' strong analytical abilities but lack of communication skills made me chuckle. The article makes many valid points - I won't quote them all here - but the take-away message is that there is always a bias toward making money in banking, in contrast with the law, where the two sides present their arguments in a balanced manner and there is not supposed to be a bias toward either side. Risk managers, well aware that they got in the way, tended to give traders the benefit of the doubt.

But risk managers should not be exonerated that easily and shift all the blame on traders. As early as March 2007, I attended a seminar in New York by Prof. Andrew Lo from MIT's Laboratory for Financial Engineering, who warned that hedge funds' computer-driven strategies would become highly correlated in times of crisis - everybody would try to sell. (See my old post here.) The seminar was organized by the MIT Club of New York and held in the Lehman Brothers building, of all places, in downtown Manhattan. Many people working in finance at well-regarded New York companies attended the talk. People laughed at Lo's joke about the implosion of the hedge fund Eifuku (which is supposed to mean "eternal luck" in Japanese, although Lo said some investors thought it was phonetic English) and agreed it was a very good presentation of Lo's research. But it doesn't look like many used his insights to check their own risk models.