Constrained Bliss Point

Yesterday was unequivocally the high point in my graduate school career to date. The big event was our first Micro 2 class, in game theory. Micro was the only class we hadn’t had yet, and my expectations were high: Econometrics is typically dry and exceedingly difficult, and our Macro class is shaping up to be intense, courtesy of our new professor. I was hoping that Micro could be the class to keep me sane this semester.

Luca Anderlini is our professor for Micro. He’s the new Director of Graduate Studies too, so my performance in Micro serves the dual role of not failing out of the program and not embarrassing myself in front of the guy running things. I had seen him present a paper last semester, and this gave me high hopes. He had a sense of humor, an entertaining manner of lecturing, and a way of making the topics at hand seem relevant.

Let me cut to the chase: my hopes were realized. The lecture was interesting, but most important, something crucial happened, something I have been waiting my entire time at Georgetown to hear someone admit. Before Professor Anderlini got into the meat of the lecture, he made a caveat. He expressed to us, in no uncertain terms, that math is not the point of what we’re doing. While, he explained, he enjoys math a great deal, and even considered a career in math, he stressed that math is a just a tool to clarify our thinking. Anyone can reason, he argued, and make a convincing case. The key is that math is a rigorous formal language to express our ideas,  so that we can make sure we are not just deluding ourselves with words. Again, math is not the end, it is only the means.

That was the breath of fresh air I needed.

Kiva.org is a person-to-person micro-lending website, which allows prospective do-gooders in the developed world to fund micro-finance operations for entrepreneurs in the developing world. I found out about Kiva around two years ago, and even though I gave a few gift certificates, it took me until today to make my first loan.

My lendee is so awesome that I felt the distinct need to blog about her. First, her name is Joice Pita, which is cool in and of itself. She lives in South Sudan, and runs a pub. I know very little about the Sudan — Wikipedia reminded me that Darfur is part of the country, and also noted that Sudan’s motto is “victory is ours” — but I can posit a guess that they could use more pubs. The thought of helping a pub-owner in the Sudan was too much to pass up.

Here’s her blurb, straight from her Kiva page:

Joice Pita is currently in the business of selling local alcoholic beverages, beer, and soda, and is requesting a loan to stock more crates of beer and soda to sell. Joice is 33 years old and is married with a husband that is a soldier. She has 6 children, and her children go to school. With the extra profits from her loan, she hopes to be able to open a hotel.

Now try and tell me that is not a cause worth funding. I thought so.

If you have some spare time, definitely check out Kiva. The money you put in isn’t a donation or a handout (though you can donate to Kiva.org itself to cover their operating expenses), which means that when your lendees pay you back, you can find new lendees and start the cycle over. You can even withdraw the money in your account after you’ve done some lending with it. So, if your bank account has some extra money in it, and you decide that instead of earning one percent interest you want to help save the world, you should head over to Kiva.org and start lending, like rite nao.

Not to be alarmist or anything, but Free and Open Source Software (FOSS) is probably to blame in bringing the global financial system to its knees. A few months ago I came across an intriguing article in the New York Times about fat tails and gaussian copulas. It was a pretty good piece, worth at least a glance.

The interesting part begins on page six. Long story short, JPMorgan developed a way of using maths to quantify financial risk into a dollar value. Value at Risk, or VaR, as it was abbreviated, was a useful tool internal to JPMorgan. Then they did something totally bonkers: they gave VaR away. Anyone who wanted to learn and implement VaR could do it, and JPMorgan would help you out. Why would they just give away such a valuable piece of proprietary technology? This quote sums it up nicely:

As Guldimann wrote years later, “Many wondered what the bank was trying to accomplish by giving away ‘proprietary’ methodologies and lots of data, but not selling any products or services.” He continued, “It popularized a methodology and made it a market standard, and it enhanced the image of JPMorgan.”

The story ends with a score of financial firms coming to rely too heavily on VaR, then they overextend themselves, get lulled into a false sense of security, and finally fat tails come in and kick everyone’s asses. Also the economy exploded.

I can’t help but wonder if it was the tactically-superior give-it-away model of FOSS that allowed JPMorgan’s mathematical monstrosity to consume the world’s financial sector in a blaze of nihilist glory.

In a previous post I challenged Georgetown’s economics department to “bring it” and kick my ass as best it could. I’m happy to report that they succeeded in winning the first round. Their tactics were a bit sneaky and underhanded, but nothing I couldn’t have anticipated. The fact is that I haven’t been putting in enough time and effort to be learning the most and getting the best grades I can. But that will change: I’m gearing up for round two.

See, what they don’t tell you about serious graduate school programs is the extent to which you are expected to know everything without having been taught it. In undergrad, the professor would only test you on things you covered in class. In grad school, you are lucky if what the professor covers in class is vaguely useful. More than half of what I’ve learned so far has been outside the classroom. This trend will surely continue into the future.

There’s also the little caveat about expectations. Some classes grade homework assignments strictly and care more about answers than methods and effort; some classes don’t even bother to grade homeworks (though if you don’t do them you’re almost guaranteed to fail the exams). Some exams test whether you can regurgitate proofs seen in class, some test mechanical problem solving skills and intuition, and some want you to know damn near everything. Sometimes you may spend more than half of your allotted time on two questions that, you find out after the fact, were only actually worth 18% of the exam grade, and you didn’t get any partial credit on them besides, but that your last five minutes of scribbling on a seemingly unimportant question netted you the majority of your points (a question which, by the way, was worth almost half of the points on the exam). No, you shouldn’t expect all exam questions to be weighted equally, either.

I know how I’m studying for the next round of exams: memorizing proofs, practicing my mechanics, and trying to learn damn near everything. And I am not making any more assumptions about how many points each question is worth.

I’m an intuitive person: I’m more concerned with broad-reaching theories than with any particular instantiation of fact. I scored a solid N on the Myers-Briggs Type Indicator. That’s not to say I don’t like facts. On the contrary, theories need to be generalized from somewhere, and starting with the known state of the world is common sense. But I don’t operate in Sensing-land — I need to be able to clearly see the general pattern in order to understand something.

Framed this way, I like economics for the same reason I like engineering. The extensive maths used in both fields are a means of grounding one’s intuition in something substantial, something provable. Maths allow one to derive, from a set of axioms and real-world data, a bunch of consistent theorems, equations, etc. that describe the system the axioms (and data) generate. If your intuition clashes with your results, either you’re wrong or you made a math mistake. Here you have a very solid check to balance out your intuition.

This allows me to satisfy two competing objectives: first, I don’t want to spend all my time in math land; second, I don’t want my intuitions to be wildly off base. So I ponder my intuition, I learn the requisite maths, and then I can go write a bunch of equations to ensure that my intuition is correct.

The internet is an unprecedented treasure-trove of quantitative data on social systems. Before now researchers could only dream of such data. I believe this provides the potentially fruitful opportunity to engage in formal, mathematical analysis of social systems (while remaining in a social-scientific vein, of course).

In the past century the field of economics underwent a mathematical revolution, focusing theoretical research on formal (mathematical) models to aid analysis and intuition, and empirical research to strengthen or rebut those models. This would have been impossible if the phenomena economics studies hadn’t been quantitative in nature; for example, if GDP numbers and employment data weren’t regularly compiled by developed nations, then modern macroeconomics could not have developed. This is one example of readily accessible quantitative data fueling a course of research, and surely there are others. (And perhaps the causation runs the other way too: when a group wants to study something, they do all they can to find good data, but this seems like the weaker direction.) Notable is the opposite case, though difficult to observe: when a particular field or direction of research is impossible because of a lack of useful data. Particular objects of study I’m interested in include “attention” or “influence,” which do not have clearly observable and measurable forms (what is the smallest unit of influence, anyway?); maybe something like “reputation”; or even abstract measures of happiness (utility, anyone?), which can explain why economics often treats consumption as a drop-in replacement. And so I turn to a study of the internet.

Out of the epic amounts of information generated by the plethora of internet-born social structures, data are lurking that could revolutionize the way we study social behavior. These data haven’t been compiled in appropriate forms yet — there is a lot of theoretical work to be done on what paradigm to use, how to aggregate the data, etc. I task Web Ecology with making sense of the morass of information, and turning it into useful data for researchers to study, build models around, and predict emergent phenomena. Once relevant data are identified and available, we can get to work on the fun stuff: the science of understanding people.

Economics has the most developed mathematical tool set for studying social behavior of any social science. With access to the data described above, researchers have an opportunity to apply those tools to something completely different, and maybe supremely worthwhile. For example, perhaps everyday social interaction (which we can approximate by interactions on the internet, maybe) can be modeled using the tried-and-true machinery of neoclassical microeconomics. Consumers of “reputational” or “social” goods would optimize utility given their (expanded notion of) budget constraints. What would be the arguments to their utility functions? What would prices and income represent in this system? All the quantities would represent something different, but all the relationships would stay the same.

Possible? Sure. Likely? Not really. But when we are harvesting the right data from the internet, we will be able to put these ideas to the test. We will be able to make changes to the theoretical apparatuses we have, develop some we haven’t even imagined, and set those to work helping us better understand human behavior.

I’m in my first year of the economics PhD program at Georgetown University, and classes started last week. Everyone I talk to says how hard the first year is. Everyone. They are so certain of the time and effort we’ll have to put in, the sixty hours a week, the studying every day in the library. I’m going to be honest and say that I don’t entirely believe them.

Of course it will be tough. Tough is what I signed up for. But economics is not rocket science. And let’s be honest, I’m kind of good at school. At this point in my life I’ve been through a lot of it, and I’ve gotten my ass kicked a fair number of times. I am certainly no n00b.

So you know what? I hope the econ department gives me its worst. I’m ready. I can take it.

I don’t mean that dating has anything wrong with it. I am thinking: How would you reduce dating down to a solvable problem? The problem itself would probably be very hard. Like you would need dynamic programming to solve it, or it wouldn’t have a closed-form solution or something.

Here’s how I would frame The Dating Problem: You are trying to find your optimal match given a set of constraints. The constraints include imperfect information, finite time, some restrictions on preferences, etc. But the imperfect information constraint is very broad. Not only is it costly to determine information about a significant other, but it is also costly to determine your own preferences.

Let’s pin things down a bit more. Imagine a game with an open-ended number of rounds. Each round is a pairing between you and someone you’re dating (ignore how you might actually find such possible significant others for now). You are trying to 1) see if they are a good (best) fit for you, and 2) you are also trying to define your preferences regarding what would make for a best fit.

That is crazy.

I think the best strategy is where, early in the game, most of your matching is geared towards determining your preferences. Then in the later stage, your rounds become shorter, since you only need to make sure you have a good (best) match. I guess the downside is you could erroneously break up with someone in one of the early stages who you would have  been better off with. But given the constraints, it’s not like you could have known, right?

And no one ever said local optima are necessarily global optima.

The economics of slavery is an incredibly interesting topic to me. More accurately, what I imagine “the economics of slavery” entails is very interesting, as I have only begun studying it, and only at an amateur level.

And what got me started thinking about the economics of slavery? Robots. Specifically the movie “I, Robot”. I’m a fan of imagining in vivid detail post-apocalyptic futures, and a robot revolution is up there with a Class 4 zombie outbreak. At the end of that movie, the overtones of Sonny freeing a generation of robot “slaves” got me thinking about the economic distinction between capital and labor (I was taking a class in intermediate macroeconomics at the time).

If we had an entire workforce of humanoid robots doing things that people usually do, what would this make the economy look like? Traditional economics would probably treat the robots as capital. But then your capital is kind of producing labor, and that is weird. The only other time that happens (that I can think of) is slavery.

I still think it’s insane that people used to own people. But, from a strictly analytical perspective, if people are owning people (and not just purchasing their labor from them), doesn’t it make sense to treat owned people as capital? There are similarities, like people tend to depreciate rather quickly if you don’t feed them and give them shelter. I read somewhere that slavery is the only time when capital and labor are equal.

I guess I’m still not convinced that there should be a tight delineation between capital and labor. Need to look into that in the future.

I think a lot about peoples’ preferences and how individuals make decisions to achieve their desired ends. Microeconomic theory touches on these sorts of questions, which I think is why I’m drawn to it. But while micro theory does some good explaining, it doesn’t go the whole way. I don’t just want a descriptive framework that maybe works good enough in most cases. I want some heuristic models that I can apply to my everyday life.

So I’m going to bastardize microeconomics and use it as I see fit. Here goes:

A rational decision maker will be better off the more malleable his preferences are. Imagine a continuum of control over one’s preferences. At one end, individuals are endowed with perfectly stable, static preferences at birth. At the other end, individuals can change their preferences however they see fit, so that only if they starve to death will their utility be less than infinite.

Now, I’m pretty sure standard utility theory presupposes stable preferences. With malleable preferences, you’re essentially optimizing two things at the same time: the mix of goods and services you purchase, and how you feel about the goods and services you purchase. That would probably lead to some pretty ridiculous maths assuming a tractable solution exists.

Behavioral economics has shown us that, no, preferences are not always pre-formed and stable. For a very cool paper on this point, try reading “Tom Sawyer and the construction of value” by Ariely, Lowenstein, and Prelec. I mean, sometimes our preferences are well-defined, like if I know I dislike strawberries. But to me the interesting case is when we are in a new situation and don’t have our preferences defined. For example: Do I like guava? I’m not sure. Allow me to assume that I do. Some other behavioral economics studies have shown that you can actually frame experiences so that you are more inclined to like (or dislike) them.

Obviously if your preferences are malleable enough, you will be happy with pretty much anything you consume. On first face this may seem like a trivial point. But I think the question “What shall I choose to consume to maximize my happiness?” is one step too far. The question instead should be “How shall I best maximize my happiness?”. Utility theory should help answer that question, I think, in as rigorous a manner as possible.

I wonder if I can formalize this idea using a neoclassical utility framework.