Thursday, April 30, 2009
Measuring efficiency in the NL, again
The motion chart/data set I've been working on is now at steady-state. I'll try to update this every Monday, not because you care, but because I'm curious about what this will look like. There are a lot of ways to figure out how well a team ought to be doing, and this method is by no means as straightforward as the simple Pythagorean theorem method based on runs scored versus runs allowed. But it does make for some very pretty pictures, and some estimates of losses due to offense versus pitching/defense (assuming all teams' have around the same variance in runs scored from game to game). There's a walkthrough in the second tab that takes you through some potentially interesting views of the data.
Wednesday, April 29, 2009
Swine flu -- now and later
When I first heard news about swine flu, I asked my girlfriend, a medical student, if I should be worried. Having just returned from Uganda where she had worked in a hospital where a resident on call could check in over 60 patients in an evening only to see all but a handful pass away before the end of that call shift, she responded with a simple, "No." Seeking a second opinion, and still worried, I found this blog, written by a Columbia University virologist. He argues that the seasonality of flu transmission will reduce the spread of infection in the northern hemisphere while tending to exacerbate it in the southern over the next few months. Apparently, flu is much harder to transmit in the hot weather of summer, creating, perhaps, the first summer where New Yorkers can be thankful for their steam bath of a city. From the blog:
Q: Should this be considered a prime candidate for next winters flu season? A: It depends on what happens in the southern hemisphere. In the next week or two we will know whether A/California/07/2009 (H1N1) spreads in the lower half of the globe and causes epidemics of disease. If it does, then it is highly likely that the virus will return here in the fall. If the virus fails to spread, then everyone can go back to worrying about H5N1.
Luckily for us, if H1N1 becomes a deadly epidemic in the southern hemisphere, we may have just enough time to produce a working vaccine by the time it attacks the northern hemisphere in force as temperatures drop. So I'm less worried now than I was, but still feel the urge to make swine flu jokes about once an hour -- a real indication of anxiety.
Tuesday, April 28, 2009
Banks complain that stress test is too fair
Regulators have told Bank of America Corp. and Citigroup Inc. that the banks may need to raise more capital based on early results of the government's so-called stress tests of lenders, according to people familiar with the situation.
The capital shortfall amounts to billions of dollars at Bank of America, based in Charlotte, N.C., people familiar with the bank said.
Executives at both banks are objecting to the preliminary findings, which emerged from the government's scrutiny of 19 large financial institutions. The two banks are planning to respond with detailed rebuttals, these people said, with Bank of America's appeal expected ...
via online.wsj.com
This article is behind a subscription wall, but I get the print version of the WSJ at school, so I got to read these final three paragraphs, true journalistic gems that they are:
One question is how the government is projecting banks' revenue streams through 2010. Some bankers are optimistic that the Fed will use their first-quarter numbers to predict their performance for teh enxt two years.
That could inflate the banks' earning potentials--and thus their capital cushions--because many of the companies had strong first-quarter performances.
Analysts, investors and most executives say those results probably aren't sustainable.
Shorter WSJ: Banks want everyone to believe they can sustain first quarter's profitability; everyone thinks that's bullshit. Given that regulators believe Citi and BofA to be undercapitalized by billions of dollars, it's unlikely that they're projecting forward the banks anomalous first quarter results, which are highly suspect as I've written about earlier. So that puts the government in agreement with, apparently, everyone else.
If only there were a "too big to lie about financial performance" corollary to the "too big to fail" maxim. I would also settle for "too big to complain about executive pay caps after destroying the economy." Come on guys, be the bigger banks.
Monday, April 27, 2009
NL Batting Efficiency - 4/27/09, Take 2
Well, after struggling to get this embedded into the post, I've given up (Verifiable -- way easier than Google to embed into Posterous, but sadly lacking a motion chart). Here's a link to a Google motion chart showing NL batting efficiency, now normalized by the number of games each team has played. As you can see, the Mets are not doing so hot. While their efficiency has increased, that's due to their getting fewer hits while scoring the same number of runs -- not the way you want to go. And their winning percentage deteriorated. Plotting runs per game on the x-axis and winning percentage on the y-axis provides a neat picture of how well teams are actually turning offense into wins.You can watch the Mets drop like a rock -- scoring the same number of runs, but getting fewer wins thanks to some truly execrable starting pitching (see: Perez, Oliver).
Thursday, April 23, 2009
Untitled
Wired loves to come up with cool ways of graphically representing data and relationships. I usually like them. But this "Enigmatrix" is bizarre.
- Game theory and Math are unconnected. Huh? Game theory is math. What do you think John Nash was scribbling on all those windows?
- Dungeons & Dragons does not connect to Plot, but Magic: The Gathering does. WTF is the plot of Magic: The Gathering? It's a money-sink that increases your social seclusion in junior high. What's the plot of Dungeons & Dragons? Ask my elven fighter-mage on a quest to restore his homeland.
- The Magic node connects to the Code node. I don't know anything about writing code or magic tricks, but I use Windows -- nothing magical about it.
- Jeopardy is in a connecting chain to Plot with Propaganda. Alex Trebek -- the Joseph Goebbels of game shows.
- Settlers of Cataan is not the central node.
Pour one out for Chrysler
Fiat denied any plans to invest directly into Chrysler, the heavily indebted US carmaker, or fund it in the future as it reported a wider-than-expected first-quarter loss.
via ft.com
Well, it was nice knowing you, Chrysler. I have fond memories of riding to Little League games in my parents' used Town & Country. That was a comfortable minivan.
Wednesday, April 22, 2009
The first time was for information. The next 182 times, not so much.
WASHINGTON — The Bush administration applied relentless pressure on interrogators to use harsh methods on detainees in part to find evidence of cooperation between al Qaida and the late Iraqi dictator Saddam Hussein's regime, according to a former senior U.S. intelligence official and a former Army psychiatrist.
via mcclatchydc.com
Remember how I was curious about how the need for 183 and 83 waterboarding sessions on two prisoners lined up with a former agent's claim that Abu Zubaydah gave everything up after his first session? Seems that those next 82 times in Zubaydah's case were not for information, but to get him to say that Al Qaedah had ties with Iraq so that we could invade. This is exactly analogous with how torture is generally used by oppressive regimes throughout history -- not to gather reliable information, but to force false confessions.
I guess that explains it.
Tuesday, April 21, 2009
Monday, April 20, 2009
The New York Mets -- the worst offense in the NL
Yes, that's a provocative statement. But consider this -- when compared to other NL teams, the Mets are the least efficient in turning offensive baserunners (as measured by hits (H), walks (BB) and batters hit-by-pitches (HBP)) into runs.
While the Mets have scored 55 runs -- more than 7 other NL teams, they are third in H + BB + HBP. Using a method of measuring efficiency called Data Envelopment Analysis and Excel's Solver function, we can score each team on the efficiency by which they turn inputs (H + BB + HBP) into outputs (runs). And the Mets are dead last in this efficiency score -- behind even the Nationals.
The chart below plots H + BB + HBP along the X-axis and runs along the Y-axis. If you were to drop a string down from the top of the graph, and imagine that each point is a peg, the pegs along which that string would lie when pulled taught are the most efficient teams. I've color-coded the points by efficiency, to make it a bit easier to see. Running your mouse over the graph reveals that the teams along the efficient frontier are the Phillies, the Marlins, the Dodgers, the Rockies, and the Diamondbacks. These teams have an efficiency score of 1. The Mets are furthest below the frontier, and have the lowest efficiency score -- 0.77. The next lowest are the Nats with a score of 0.80.
Obviously, it's not all about runs. And it's not all about efficiency -- if the Mets were collecting 50 hits a game to score 10 runs, nobody would be complaining, even though they'd be horribly inefficient. But what's so worrisome about the Mets is that it's not like people aren't hitting -- the team's not slumping, in the usual sense, and yet they still can't generate enough runs to win consistently. This crude measure of efficiency reveals exactly what Met-fans have been griping about -- the Mets are leaving a lot of runners on base. I didn't realize just how bad it was.
He spilled it after 35 seconds ... and then we waterboarded him 82 more times
With the recently released torture memos making it clear just how many times we waterboarded Abu Zubaydah and Khalid Shaikh Mohammed (83 and 183 respectively), my mind raced back to a vaguely recalled article from a couple years ago. In this article, there was a statistic about how quickly waterboarding had worked to loosen the terrorism suspect's tongue, and I remembered being simultaneously horrified and impressed with the speed at which it worked. A little digging, and I came up found the article.
An AP article that quoted a former CIA agent from December of '07, it was one of the first public confirmations that we had used waterboarding. Interesting, though, is the fact that the former agent claimed Abu Zubaydah was talking "in less than 35 seconds." We now know from the memos that that's at the long-end of a normal waterboarding session. So, if we can believe this agent, they used it once to great effect, and then used it 82 more times.
So, one of two things is true.
1. It was really effective and we tortured him 82 more times anyway.
2. It wasn't effective, but we tried it 83 times.
Which one nauseates you more?
Saturday, April 18, 2009
Further explorations of CEO compensation
I was pointed to the data visualization website Verifiable by this blog post. I then came across a chart of total CEO compensation against stock performance, showing little correlation between performance and pay.
I snagged the data and made some changes. Since stock performance is highly variable, and CEOs probably don't have all that much control over one-year stock returns, I was curious to see if CEO compensation was related at all to the size of the company. After all, it's reasonable that boards of directors will consider the difficulty of CEO's job to increase as the company grows larger and its operations more complex, and so pay more to entice top talent. Market cap may not be the best proxy for size and complexity, but it's pretty good.
I excluded Apple, where Steve Jobs makes $1 a year, and took the log of both total compensation (including stock options and grants) and total cash compensation, and compared them with the log of market capitalization. As you can see, there does some to be a weak correlation in both charts. The correlation between total compensation and market cap is 0.36 and that drops to 0.33 between cash compensation and market cap.
<a alt="verifiable.com" title="verifiable.com" href="http://verifiable.com/charts/2684"><img style="border:0;max-width:800px;max-height:600px;" src="
" /></a>So what have we learned? Not that much. There's probably an extensive academic literature on this subject that I haven't looked at. But this does suggest that CEO compensation isn't a total black box.
Friday, April 17, 2009
Stanky banks
Go Citigroup! Best quarter since 2007! Thanks to an accounting trick and revenue from their fixed income trading unit.
What is with these fixed income trading units? They made JP Morgan money, they helped Goldman sell $5 billion in stock. In fact, they seem to be single-handedly keeping their banks in the black. But what are they trading, and why is it they're making so much money now?
Seems that AIG, owned by you and me, is selling these banks entire portfolios of fixed income instruments at really, really low prices. This is what it means, apparently, to "unwind" AIG's business -- sell still-valuable assets to investment banks at well below their market value in portfolio-sized transactions. The investment banks make big profits, and the soon-to-be-unemployed managers of these trading desks at AIG are eminently employable heroes to the investment banks.
But how could this happen? Probably because the government wants it to happen. This is one sneaky, anti-capitalist, anti-markets, anti-democratic method of giving more money to ailing investment banks. With Congress unwilling to give more money to bail the big banks out, these transactions provide another injection of capital into the banks through a third-party -- AIG.
One more quick observation: As Obama delivers noble rhetoric about why torture can't be justified in the name of expediency, his Treasury Department is subverting free markets and the democratic system to save the banking sector. I'm not drawing a moral equivalence, or suggesting that saving the financial system isn't necessary. Call me crazy, but funneling money to i-banks through selling AIG's tax-payer owned assets for cents on the dollar does seem logically analogous to torturing people for the sake of potentially life-saving information; in both cases, you damage our democracy for the sake of expediency. No caterpillars in bank vaults yet, though.
Thursday, April 16, 2009
Who to pay and how much, or CEOs don't earn what they make
This is really the critical point. According to the standard textbook model, an increase (or a decline) in the labour income of a given individual should be interpreted as a rise (or a fall) in his or her marginal product, i.e. his or her contribution to total economic output. That is, if Mr Smith's wage rises from $30,000 to $50,000, then it must be that Mr Smith has produced $20,000 of extra economic output. The beauty of the market system is precisely that the increase in Mr Smith's wage should in principle correspond to the creation of new economic value and well-being, and is not obtained at the expense of anybody else (i.e. even with fully selfish economic agents there is no externality on others that is not being internalised by the price system). This textbook model probably provides an (approximately) accurate description of 99% of the labour market. However it is extremely naïve—to say the least—to imagine that it adequately describes the pay determination process at the very top end of the labour market. Assume that the CEO of AIG or GM manages to get an increase in compensation, say a rise from a $5m to a $10m total annual compensation. It is truly heroic to conclude from this observation that his or her contribution to AIG or GM output has increased by $5m, and that the total output of AIG and GM has risen by that much. There is tremendous evidence showing that the invisible hand of the market simply does not work in this very peculiar segment of the labour market, and that top executives will keep setting their own pay to the highest possible levels (with no connection whatsoever with their marginal product, which nobody can properly estimate) as long as they are not prevented to do so. Historical evidence suggests that highly progressive taxation on very high incomes is the most efficient way to achieve this goal.
via economist.com
This from Professor Thomas Piketty in his Economist debate with Cato Institute's Chris Edwards about instituting high (like 80% marginal tax rates) on incomes over a couple million dollars. It's a good summation of the back and forth that's been happening among policy wonks over how to regulate executive pay.
In this paragraph, Piketty puts his finger on the central perceived problem with paying CEOs a lot of money -- they don't produce value equal to what they earn. But what I haven't seen anywhere is much discussion of the reasons for setting salaries beyond a worker's marginal product, and there are a lot of other reasons.
Most relevant for CEO pay is the way someone else's pay -- someone above me in the corporate hierarchy -- can affect how hard I work. When I am making a decision about how much effort to expend in the current period, I'm balancing the cost of my effort against the benefit I get not only from my current salary, but also from the improved probability of being promoted, and the concomitant salary boost. So if I'm an Executive Vice President at a Fortune 500 company competing with five other EVPs and innumerable outside candidates to become the next CEO of the company, the CEO pay package could incentivize me to work harder to win that promotion. Without that incentive, you may have to pay me more. Paying me and my five other competitors more may actually cost you more than what you saved by paying the CEO less. And it's not that you're paying the CEO for her work -- some of her pay is just to incentivize me.
Obviously, this effect may not explain all big CEO pay packages, but it's at least worth thinking about, because if you don't, you're really not thinking about the problem in a complete way, and could end up misclassifying good corporate policy as bad, or criminal.
To me, big CEO pay packages are not a root problem, but a symptom of weak corporate governance. If we could be reasonably certain that a company's board was truly independent and their CEO was still making a ton of money -- more than they earn themselves -- then we could be reasonably confident that the company had other good reasons, like providing incentives to others within the organization.
Tuesday, April 14, 2009
Goldman Sachs thinks we're stupid
6:50 a.m.| Where’s December?: Goldman Sachs reported a profit of $1.8 billion in the first quarter, and plans to sell $5 billion in stock and get out of the government’s clutches, if it can.
How did it do that? One way was to hide a lot of losses in not-so-plain sight.
Goldman’s 2008 fiscal year ended Nov. 30. This year the company is switching to a calendar year. The leaves December as an orphan month, one that will be largely ignored. In Goldman’s earnings statement, and in most of the news reports, the quarter ended March 31 is compared to the quarter last year that ended in February.
The orphan month featured — surprise — lots of write-offs. The pretax loss was $1.3 billion, and the after-tax loss was $780 million.
Would the firm have had a profit if it had stuck to its old calendar, and had to include December and exclude March?
How to run a profitable financial institution in five easy steps.
1. Shift the fiscal year so that it excludes a bad month (December) and includes a better one (March).
2. Stick all your write-offs in the bad month.
3. Report a big quarterly profit.
4. Sell $5 billion of stock to the suckers who believe you.
5. Pay back the government so you can give yourself that big bonus you deserve.
Lloyd Blankfein, everybody. It's also worth noting that Goldman Sachs has set aside $4.8 billion for salary and bonuses -- a larger percentage of revenue than it set aside last year. Also a number suggestively similar in size to the $5 billion in stock they want to sell.
I hope to God this doesn't work out for them.
Tuesday, April 7, 2009
Why shouldn't I buy one of these?
Pogoplug review
by Joshua Topolsky, posted Apr 6th 2009 at 12:27PM
When we first caught wind of the Pogoplug -- a small box that essentially lets you turn any USB hard drive (and drives only) into a network device -- we were pretty darn excited. Having a house full of disparate storage boxes and no easy way to connect to them made the prospect of the 'plug seem very enticing. Not only does the Pogoplug make your drive accessible via your PC (with accompanying software), but it -- we think more importantly -- makes the drive accessible via a web front-end and an iPhone app. We finally had a chance to break one of these out and see how it performs, and our findings are below.
via engadget.com
Sounds pretty sweet. And I don't use my external HD at all as it stands.
Monday, April 6, 2009
I am not as fun as I think I am
While procrastinating by using my monthly emusic downloads today, I realized that I really don't understand my own music consumption very well. Or at least my purchasing decisions do not match my consumption habits. Here is a chart detailing the degree to which I over/underweigh my music purchasing by type of consumption:
I know I listen to music a lot while studying and working. So I download a lot of jazz, electronica (i.e. Boards of Canada), and indie music where vocals are fuzzy (i.e. Sea and Cake, Beach House). But I also download music because I would love to listen to it at a party. But I host 0 parties. And I'm still waiting for that first invitation to DJ a party.
Subscribe to:
Posts (Atom)