Month: September 2009

Models, Information, and Market Rationality

I have come across a couple of items recently involving market rationality and the ability of the market to reflect “unknown” information.  The first came in a conversation with my colleague Bryan Routledge.  Harkening back to the Challenger disaster, Bryan mentioned that “the market” quickly determined the company that caused the failure (all this is my paraphrasing of my understanding of what Bryan said:  it is my fault if it is wrong!).  Here is the graph of the stock prices of the main companies involved in building the space shuttle on the day of the disaster:

shuttle company stock prices

There are four companies shown:  Morton Thiokol, Lockheed, Martin Marietta, and Rockwell.  The stock price for all of the companies immediately dropped 7-8% after the disaster.  Within an hour, three companies went back up to being just 2-3% down, while one company further decreased:  Morton Thiokol.  The company responsible for the O-ring (of Richard Feynman and ice water fame):  Morton Thiokol.  It is certainly provocative that the market seemed to know something immediately that took an investigation months to determine. It would have been even more impressive if the market identified this an hour earlier (the explosion happened at 11:39AM), but the results from the day are still pretty impressive.

But, as Bryan reminds me, this was not exactly a mystery to everyone at the time:  the engineers involved strongly suspected early what the issue was and later fed that information to Feynman.  So the information was out there and perhaps that information leaked out to the market in the immediate aftermath of the explosion. So perhaps it is not so mysterious after all. And there may well be other explanations for the larger drop off by Motton Thiokol.

Continuing the theme of markets seeming to have information that mere mortals do not, Panos Ipierotis, with tongue-in-cheek, suggested a prediction market on the P=NP question, arguing

So,if P=NP is a decidable problem, it is either true or false. So, a fully rational agent, participating in the market, should know whether P=NP. It is not a matter of probabilities! All the information to make the decision is available. So, if the market has one or more rational players, the market should converge to a price of 0 or 1 immediately, depending on the state of the problem. Right?

He then offers a few choices when the market, presumably, does not give such a result:

  • There are no rational agents. So, all the analysis of prediction markets that assume rationality of traders is incomplete.
  • There are rational agents. The market does not converge to 0 or 1 because the P=?NP problem is undecidable.
  • There are rational agents but the return from the risk-free rate until reaching the time to settlement exceeds the return from the market. So, the market gives information on how long it will take for the problem to be officially solved.
  • If your laptop cannot find the solution, neither can the market.

These are not mutually exclusive:  it could be more than one.  But I think it is pretty clear that there are no rational agents by this definition, and I think most or all economists will agree to that.  The concept of having a rational agent is no different than the assumption of linearity and divisibility and so on in linear programming.  A model cannot include everything (a map that includes everything would be as large as the area mapped), so simplifications are made.  Some economic models include limits on rationality, or on information, or on timing, while others will give their agents more power/information/etc. than could occur in practice.  There is normally a tradeoff:  with more detail on rationality may come limits in other areas, such as number of agents, or complexity of markets.  Given some set of assumptions, it is pretty easy to come up with a situation that exploits weaknesses in the assumptions. If that is the case, those are the wrong assumptions to make!

So I would agree with Panos: “all the analysis of prediction markets that assume rationality of traders is incomplete”.  That “incomplete” analysis might still be useful.  To paraphrase the statistician George Box: all models are incomplete;  some are useful.

This concept was certainly seen by Feynman in the Challenger case when he questioned whether some of the mathematical models used were useful.  From Feynman’s appendix to the Challenger Report:

When using a mathematical model careful attention must be given to uncertainties in the model.

Best Place for Undergraduate Engineering?

No, I am not going back to get a bachelor’s degree (actually, only my graduate degrees are in industrial engineering:  my bachelor’s is in math and computer science, so maybe I should go back!).  The son of a colleague of mine is planning to take engineering and wonders where to go.

In my mind, one of the main purposes of a university education is to get you fired up about some topic.  This “firing up” is kinda like lightning:  it is hard to tell when it is going to happen.  But enthusiastic faculty and an well-thought-out (and innovative) curriculum go a long way in improving the odds.

Is there any university that you have seen that you have really thought:  boy, that would be a great
place to get an undergraduate engineering degree?  While my colleague’s son is mainly interested in US or Canada, east of the Mississippi, I’d be happy to hear about any place for comparisons sake.

Show Off Your Best Work in Operations Research Practice

I am a huge fan of the Franz  Edelman Award for Achievement in Operations Research and the Management Sciences (best work in operations research practice) given by INFORMS.  The applications are uniformly inspiring and the presentations go way, way beyond the norm for our field.  The full papers, published every January in Interfaces, are ones that I actually look forward to (something I don’t do for my own papers!), and form a big part of an MBA course I teach here.

Being an Edelman finalist is a tremendous commitment:  in addition to the full paper, the presentation generally requires the cooperation of a Cxx of the firm  (for suitably high xx: EO is great!).  Don’t try to get by with half-baked work here:  you won’t get past the initial phase.  But if you become a finalist (let alone a winner), the fame is worth it!  This is a great opportunity to get the attention a level or three higher than you might otherwise (and give the Cxx the opportunity to brag on your behalf).

If you are doing operations research that is truly changing how an organization works, I strongly encourage you to enter it in the competition.  The initial phase only requires a 2-3 page description.  See the full details here.  Deadline is October 21, so get typing!

And the Winner of the Netflix Prize is …

BellKor’s Pragmatic Chaos!  While The Ensemble nipped BPC at the wire for the public test sets, the BellKor team did better on the hidden test set.  From the announcement:

Team BellKor’s Pragmatic Chaos edged out team The Ensemble with the winning submission coming just 24 minutes before the conclusion of the nearly three-year-long contest.  Historically the Leaderboard has only reported team scores on the quiz subset. The Prize is awarded based on teams’ test subset score. Now that the contest is closed we will be updating the Leaderboard to report team scores on both the test and quiz subsets.

Great work by all of the teams:  I think we really do understand a bit more about this data and about data mining in general due to this contest.

Looks like this was successful enough for Netflix to plan a second contest.

The True Measure of a Person’s Real Worth

Inspired by Jon Cryer‘s comments on (finally) winning an Emmy last night:

You know, I used to think that awards were just shallow tokens of momentary popularity but — I realize they are the only true measure of person’s real worth as a human being.

I thought I would ‘fess up about an award or two coming the way of yours truly (“Aaah, shucks, you shouldn’t have!”).

First, the Faculty of Mathematics at the University of Waterloo has tapped me for their 2009 Alumni Achievement Medal.  While I only spent two and a half years as an undergraduate at Waterloo (I had spent two years at the University of Manitoba), those years were incredibly important to my path.  I got my B.Math. degree  from two departments:  Computer Science and Combinatorics and Optimization.

edmondsYou have to love a place with a Faculty of Mathematics and a 29 faculty strong Department of Combinatorics and Optimization.   I was particularly attached to C&O and was inspired to enter the field of operations research due to the training I got there.   I was strongly affected by two faculty members on the opposite sides of practically any spectrum.  The first was Jack Edmonds, who founded combinatorial optimization with his work on matchings and his recognition of the key divide between polynomial and exponential algorithms.  While that work was done in the 1960s, Jack has remained a prolific and influential researcher to this day.  From Jack, I learned the beauty of this field.

The second person I was strongly influenced by will be less familiar to most:  Richard N. Burns.  Rick did research primarily in nurse scheduling.  We worked together two summers on my first operations research projects.  The first project was for Domtar, a Canadian paper products company, on scheduling a cardboard cutting machine.  For this project, I got my first computer (this was 1980 or so):  a CP/M based system.  It was a fantastic project that taught me both the pleasures and the frustrations of doing real-world operations research.    Exhilarating when things went well; devastating when the code just wouldn’t work.  But finally we got things to work well and the system went into use.  In retrospect, I now see that the work we were doing could be seen as an early precursor to branch-and-price, a topic that has fascinated me ever since.

While the Domtar project was a research project, the second project I did with Rick was a consulting project, putting together a nurse scheduling system.  By the time we had added in time tracking, accounting, payroll, capacity planning, and a few other things, I learned the definition of mission creep:  I am not sure that system ever got finished!  But that too was a great experience.  I enjoyed working with Rick very much, and from him I learned the pleasures of applying operations research in the real world.  I’ve lost track of Rick (he seems to have retired from academia, so I don’t even have a photo of him:  if any of you do, please pass it along!) but he continues to have a strong effect on me.

As you can imagine, Rick and Jack were quite different in their approaches to the world of operations research (if you know both of them, you know they were different in lots of other ways too!) but when it came time for a recommendation of where to go to graduate school they both, independently, suggested the same place: Georgia Tech Industrial and Systems Engineering, which turned out to be the ideal place for me.

I am thrilled that the Faculty of Mathematics picked me for their Alumni Achievement Medal.  The list of past recipients is very impressive indeed!  I pick up the medal at a banquet on Thursday which Alexander and I will drive up for (Ilona being in Europe at the moment).

The second award is … well, maybe I better wait until the INFORMS San Diego Meeting to talk about that.

Finally, though not really an award (perhaps a penalty might be a better term), I am now the Associate Dean, Research here at the Tepper School.  Given the research traditions of this school (and not just in OR!), I am really pleased to be part of making a great school even better.

Become Famous by Winning the COIN-OR Cup!

COIN-OR_CMYK-296Do you use COIN-OR (open source software for operations research)?  According to the log files, lots and lots of people do!  If you are doing something interesting with COIN-OR and are planning to attend the San Diego INFORMS Meeting, I strongly encourage you to enter the COIN-OR Cup!  You could join John Forrest, Jonathan Eckstein+Bill Hart+Cindy Phillips, John Tomlin and the team from the Environmental Protection Agency + Sandia National Labs (including Jonathan Berry) as a winner of this prestigious Cup!  Fame and drinks will be yours forever!  Despite appearances, you do not need a John or Jonathan on your team to win, though it seems to help.

More seriously, if you have used COIN-OR in a neat application (or have advanced COIN-OR in a significant direction), it would be great to hear about it through an application for the award.  This would let the COIN people know of all the great ways the software is being used and help bring together the COIN-OR community.  The deadline is October 2, so don’t delay.  And thanks to IBM for sponsoring the reception (which is always a real highlight for the whole COIN-OR community).

Operations Research Running Countries

In the wake of the new Japanese Prime Minister having a doctorate in operations research, David Curran (iamreddave on twitter) pointed out that there was a previous effort of having a country run by operations research (or operational research, in this case).

StaffordBe</li> </ul> <p>erStafford Beer was one of the founding people in British operational research. He was one of the people who saw operational research in World War II and adapted those methods to work in practice, in his case at United Steel, followed by some consulting companies. He ended up founding many aspects of systems science and “cybernetics” (a term I rarely hear these days). The (British) Operational Research Society gives out a medal in his honor:

Stafford Beer Medal

This award is named in memory of Stafford Beer, a world leader in the development of systems ideas, especially management cybernetics, and President of the OR Society 1970-71.

The Stafford Beer Medal is awarded in recognition of the most outstanding contribution to the philosophy, theory or practice of Information Systems and / or Knowledge Management published in the European Journal of Information Systems or Knowledge Management Research & Practice within the relevant year.

In 1970, Beer got hooked up with Salvador Allende’s Chilean government (an elected, socialist one) and he had the great idea of hooking up offices all around the country through unused telex machines. By setting up these telex machines in factories around the country, linked at a central control room, Allende’s ministers could guide production and distribution at a country-wide level.

Cybersyn_control_roomThe system, named Cybersyn, had its greatest success in reaction to a country-wide strike. There is a fascinating article from the Guardian a few years ago on this whole episode:

Across Chile, with secret support from the CIA, conservative small businessmen went on strike. Food and fuel supplies threatened to run out. Then the government realised that Cybersyn offered a way of outflanking the strikers. The telexes could be used to obtain intelligence about where scarcities were worst, and where people were still working who could alleviate them. The control rooms in Santiago were staffed day and night. People slept in them – even government ministers. “The rooms came alive in the most extraordinary way,” says Espejo. “We felt that we were in the centre of the universe.” The strike failed to bring down Allende.

There is also a wikipedia page on Cybersyn.

The story has a sad ending, of course. A military coup took over, assassinated Allende and destroyed Cybersyn. Beer, out of the country at the time, was clearly affected:

Soon after the coup, Beer left West Byfleet [his home at the time], his wife, and most of his possessions to live in a cottage in Wales.

It is a great story, and I highly recommend the Guardian article (and thank Dave for the pointers).

I both love and fear the idea of operations research running a country. There are many times I look around and think “If only we could get organized, we could run this much more efficiently”. I particularly think this when waiting in line or looking at a set of transportation schedules that don’t interact correctly.  But, given the trouble I have getting something like a sports schedule together, I don’t think having an OR model run a country is particularly realistic.  Cybersyn did show the value of having free information flow, and the coup’s response showed how threatening that flow can be to entrenched interests.

INFORMS Needs Writers

I recently had an exchange on twitter on why the OR community is not more effective on using twitter, facebook, and so on to get the story out. People like Laura McLay, Aurelie Thiele, and many others listed on the sidebar do have blogs and many of us twitter and facebook away, but we are a pretty small group. As I replied in Twitter, “We have lots of stories, but not enough storytellers”.

If you are a student (or recent graduate) have wanted to do some writing about operations research, INFORMS is ready to give you a chance, and will even provide some walking around funds. From an email making the rounds:

INFORMS

Writer for the INFORMS Annual Meeting Daily E-News: October 11-14, 2009.

INFORMS seeks 3 OR/MS students or graduates to cover on-site the sessions, events, and breaking news during the 2009 INFORMS Annual Meeting to be held in San Diego, CA, October 11-14, 2009. The writer is expected to produce content for a daily electronic newsletter throughout the duration of the 4-day meeting. The writer may have experience regularly publishing news articles in professional or collegiate news outlets or equivalent experience.

Qualifications:

– Current or recent enrollment in a college-level OR/MS program and/or equivalent experience is required
– Must be proficient with Microsoft Office (Word and Excel)
– Laptop with wireless capabilities is required
– Experience with digital photography and ownership of a digital camera is required
– Must be able to attend the conference all 4 days

INFORMS will provide a daily stipend of $100 and reimbursement for local transportation/parking cost if necessary. Please forward a resume and two writing samples (approximately 500 words each) to Ms. Mary Leszczynski, Managing Editor, mary.leszczynski@informs.org. For more information on the meeting, please visit: http://meetings.informs.org/SanDiego09/.

Operations Research Key to MBAs

Stacy Blackman, in the blog “Back to B-School”, has a short summary of the ideas of Matthew Stewart, author of The Management Myth, a book highly critical of of the world of MBA education. Since I primarily teach operations research in the Tepper MBA program, I was heartened by Stewart’s views:

While Stewart believes that highly specialized studies in areas such as process-oriented, operations research can be useful training for managers, it’s the case-study oriented, generalist programs such as Harvard Business School that are less useful. Stewart says this is a problem of content:

In order to produce generalist courses, business school professors have been forced to invent subjects called strategy, called organizational behavior, and so on. They’re pretty much pseudo-sciences, and when you use them as a basis for instruction, you’re really teaching people how to master arcane jargon that has minimal connection to the real world, as opposed to teaching them to really think.

…
Stewart would like to see MBA programs focus not only on business but on broader subjects that would be useful to developing knowledge and critical thinking, such as political theory or evolutionary biology. At the same time, he believes greater specialization is key. “Forget all this nonsense about general case studies and teach how logistics operations work in a complicated supply organization. Give them a real specialization as opposed to a phony one,” says Stewart.

I wouldn’t go as far as Stewart, at least as projected in these quotes, since I do believe there are useful insights from strategy and organizational behavior, primarily through the more formal, less “war story” teaching that you get at some business schools (including the Tepper School).

Maybe we can see a resurgence of operations research in business schools!