Archive for August 2010
Whenever I describe my trip to India this past spring, I tell people about visiting the Sikh temple in Delhi. It’s hard to describe in words, but this video does a nice job of summing up what is interesting about these temples: a massive free meal program to people of all faiths, the pool of water, and the pride the Sikhs take in rejecting the Hindu caste system. This is not the temple I visited, but it has many of the same features. Go read about Sikhism on wikipedia.
A new working paper (pdf) by some prominent economists uses a unique education experiment in the 1980s in Tennessee to find that kindergarten class size and teacher experience have significant impacts on “adult outcomes” – among other things, the future earnings of the student about 25 years later.
The experiment randomly assigned kindergarten students to either small classes (~15 students) or large classes (~22 students). Most of the research performed on these data focus on test scores. The previous studies found a “fade-out effect” – the measurable test score bumps that students in small classes received in kindergarten reduced to near zero by later grades. This new study is surprising because it finds a significant effect that re-appears in adulthood. The small-class students were more likely to go to college, marry, purchase a house, contribute to a 401k, and they earned slightly more. Remember: this is kindergarten we are talking about.
Interestingly to me, the earnings bump is not very large for small classes over large classes. The “net present value” of the effect is about $11k per student. Not too shabby, but small compared to the almost $10k spent per student to reduce the class size in the first place. A much larger future earnings effect is found to be caused by teacher experience (and teacher “quality” as measured by test scores at the end of the kindergarten year). Their rough calculation shows that an extra year of teacher experience translated to roughly $62 of extra income per student per year – or a net present value of $34k for a class of 20 students. One might argue that by not giving kindergarten teachers annual raises of $34k, society is extracting a rent from these teachers – they are providing more long-run benefit than we are willing to pay them.
The other “explosive” result is that an above-average teacher is worth a lot more to society than a below-average teacher (as measured by kindergarten test scores). The 75th percentile teacher will increase the net present value of the 20 students in their class each year by $320,000 over the same class lead by a 25th percentile teacher.
(For a primer on the IB diploma, see Wikipedia.)
After reading the interesting IB Statistical Analysis over at Wandering Academic, I was left with a number of unanswered questions. The post there deals mainly with the question of how distributions of scores have changed (or not changed) over the past five years, as well as pointing out that some courses have very different distributions than others. If I were a student, here is what I would want to know:
- Which courses and subject groups yield higher scores?
- Which extended essay subjects will maximize my chances of getting extra points?
I have answered these questions, to a certain extent, using Wandering Academic’s data. All data are aggregated 5-year historical world averages, which smooths out minor shifts in distributions over time. Only subjects in the WA dataset are included – for example, group 1 consists only of English A1. Scores for courses range from 1 to 7. Extended essays (see further below) are given grades of E (1) to A (5). Some of my numbers will differ slightly from WA’s due to rounding errors. (Also, I think the global average TOK score is incorrect in the WA dataset – according to my calculations, it is 3.24 rather than 2.76.)
Here is the full list of courses, ordered by their 5-year average scores:
Scores vary widely. Spanish B HL has an average score of 5.51, more than a point and a half higher than the average Theater SL score of 3.83. Groups 1, 2, and 3 dominate the top half of the list, and groups 4, 5, and 6 fill out the lower half.
To make this last point more clear, let’s look at things aggregated by subject group. There are six subject groups – students have to take courses from all six (actually, there’s a way to skip group 6 – more on that later) and split between HL (“higher level”) and SL (“subsidiary level”) versions. Extended essays will slot into one of the six groups, depending on the topic selected by the student.
English A1 takes the top SL and extended essay (EE) score averages, while “second language” HL courses are the highest-scoring. Group 4, experimental sciences, gets the lowest average scores across the board.
“Extra” points are interesting twist. To graduate with the IB diploma, a student has to accumulate a number of points across each of their courses. They can get up to 3 bonus or “extra” points by getting a good score in both their extended essay and the core Theory of Knowledge (TOK) class. Since the extra points are jointly determined by EE and TOK scores, as described in this matrix, I computed the expected extra points by taking the cross product of the scoring matrix and the probability matrix described by the actual distribution of scores in each group’s extended essay bucket, and TOK scores. This is a poor way to compute the expected score, actually – it assumes that the proportion of “E”-grade extended essays is the same in the group of students that get an “A” in TOK as it is in the group that gets an “E” in TOK. Obviously, this is false – they are not independent. Nevertheless, without access to the data on individual students, I guess we can consider this a more or less decent proxy for a student trying to choose which subject group to select for their extended essay, assuming they are equally interested in all subject groups.
To round out the analysis, I asked the question whether HL courses are better- or worse-scoring than SL courses (not including ab initio and Math Studies).
On the whole, HL courses score about a half-point higher than SL courses (perhaps reflecting the fact that students self-select), with the exceptions of English A1 (-0.21), History (-0.14), Biology (-0.08), and Math (-0.13). In these four subjects, it seems that the difference in difficulty of the material outweighs the self-selection effect.
On Friday, I reached a milestone of sorts at my local Starbucks: the barista retrieved my preferred drink without me having to ask for it. A small delight, not quite as great as asking for “the usual,” but close.
In my experience, barista-customer relationships at coffee shops evolve like so:
- Default: the normal interaction of ordering, receiving, and paying; small talk rare.
- Mutual Recognition: barista and customer recognize each other and treat one another with slightly more familiarity; occasional small talk; barista can sometimes anticipate the order, but still requires confirmation from customer.
- The Usual Order: barista connects usual order to customer’s face – no confirmation required; possible variation: “Which will it be today: a latte or a cappuccino?”
- Name Recognition: barista and customer know each other’s first names; frequent small talk.
- Free Stuff: barista occasionally (but regularly) declines payment from customer; in some cases, social interaction between barista and customer occurs outside the coffee shop.
Coffee shops display this unique progression for a number of reasons: the habitual nature of coffee drinking (causing regular and frequent interaction), the inexpensive nature of the product (giving baristas occasional discretion in dispensing free goods), and the fact that the interaction typically happens in the morning (pre-work) or post-work evening (creating a bond between two souls – barista and customer – commiserating over the work they both hate). Most people never get past stage 1. Reaching stage 5 requires an investment of time and energy that few are willing to risk.
Over the years, I have been both a (profligate) stage 5 barista and a stage 5 customer, each with their own benefits. As a current stage 3 customer, I enjoy a bit faster and friendlier service, but nothing of much value. Stage 5 is the key: it brings real, measurable wealth.
The funny thing is, businesses figured out the “free stuff” fetish we customers have a long time ago, especially coffee shops. They even figured out how to profit from it: it’s called price discrimination. There’s nothing sinister about this kind of discrimination. It just means that businesses use clever means to charge a higher price to people willing to pay the higher price and a lower price to people who aren’t. Case in point: the free coffee card. Buy ten drinks (and collect ten stamps) and get one free. The people willing to go through the hassle of collecting ten stamps are clearly more motivated by the free stuff fetish – so they throw us a bone.
Starbucks has a different twist on this kind of loyalty program. Their “gold card” is a stored value card. You use a credit card to load money onto it, then for each transaction you pay for using the gold card, you get a star. For every 15 stars you collect, they mail you a coupon for a free drink. The brilliance of this system is that the stored value card transactions do not incur a credit card processing fee – they literally save a dime or more every time I pay them using their silly card instead of with my Visa debit card. They also get interest income on the balance that sits in my “account” before I get around to buying coffee with it. So everybody wins: they get a loyalty program that more than pays for itself and I get free stuff every once in a while.
If you dig around in the official Starbucks annual filings, you’ll find some other interesting tidbits. For example, Starbucks made $26 million in 2009 on card values that are unlikely to be redeemed – i.e. the 25 cents left on a card that hasn’t been used in 4 years, or whatever the rule is. The rule used is, of course, “determined by management.” Twenty-six million dollars is double the amount from the previous year. (This is likely one of their “cookie jar” accounts that Starbucks management can legally manipulate to beat their target earnings numbers. That’s another post for another day.) Total “deferred revenue” (which is where Starbucks card balances show up) amount to about $389 million.
Other businesses offer similar loyalty programs. Frequent flyers of particular airlines collect points, which they can use for free flights or, increasingly, merchandise or gift cards. Even debit cards are in the on points game these days. And these points have value – they can be exchanged for valuable things, after all, much like currency. The key is that airlines can control very carefully how and when the miles can be exchanged. Whereas Starbucks simply gives you a free drink of your choice, airlines institute black-out dates during which you cannot redeem miles for travel and can change the “price” of goods at a moment’s notice.
I can’t verify the numbers in this table, but they are roughly equal to other quotes from around the interweb. They show that the total outstanding number of airline miles was over 14 trillion (that’s a “t”) back in 2005 and growing at a nice smooth exponential rate. For a standard domestic 30,000-mile ticket, 14 trillion miles represents more than 450 million free tickets.
All that loyalty is starting to look costly. These airlines are carrying a huge liability (something they owe) in the form of airline miles. They wold love to reduce it somehow other than giving out all the free tickets owed. According to United Airline’s 2009 annual report, they did just that back in 2007 – reaping $246 million by simply changing the “inactive account” limit from 36 months to 18 months (see footnote (a) on page 44). United claims a total of $4.2 billion in “Mileage Plus deferred revenue” (page 76) – basically, the value of all those miles on the books, that they someday owe their customers for being loyal. This liability is more than the total shareholder deficit of $2.8 billion. In other words, if not for the frequent flyer program, United Airlines might be worth something.
In the end, we continue to chase free stuff by paying for it. In the meantime, I get my morning coffee without lifting a finger, and feel for a moment, well, important.
Speaking of outlandish politicians, Wyclef Jean announced his candidacy for the presidency of Haiti.
I love Wyclef’s music. As this journalist points out, he is “not the most naturally gifted musician,” but I have always dug his zany sentimentality. See, for example, “911″. And the Kenny Rogers bits on Ecleftic.
I have a hard time imagining this translating to effective national leadership, however.
Jon Gnarr, comedian and leader of Iceland’s “Best Party”, is the new mayor of Reykjavik, with promises of free towels at public swimming pools and a new polar bear exhibit at the zoo. The Best Party won 6 of the city’s 15 council seats. They needed to form a coalition, but “ruled out any party whose members had not seen all five seasons of ‘The Wire.’”
During his acceptance speech, he sought to assuage concerns that he wouldn’t take his job seriously:
No one has to be afraid of the Best Party, because it is the best party. If it wasn’t, it would be called the Worst Party or the Bad Party. We would never work with a party like that.
From the Christian Science Monitor comes this beautiful headline:
Monkeys hate flying squirrels, report monkey-annoyance experts
Japanese macaques will completely flip out when presented with flying squirrels, a new study in monkey-antagonism has found. The research could pave the way for advanced methods of enraging monkeys.
Via Marginal Revolution.