Wednesday, December 17, 2014

Fact Checking

Recently New York Magazine published an amazing story about a Stuyvesant High School student who had purportedly  made $72 million in the stock market.  When I read it I was pretty sure it was nonsense as in fact it soon turned out to be.  As the magazine admits its fact checking was obviously inadequate.  I expect this was caused in part by a failure to realize just how unlikely this story was and the implications for the appropriate level of fact checking.   Suppose for example that there are a million false claims like this for every true account.  Then fact checking sufficient to catch 99.9% of the false claims is totally inadequate as 99.9% of the surviving claims will still be false.   As the saying goes "Extraordinary claims require extraordinary evidence".

New York Magazine stated:

... As part of the research process, the magazine sent a fact-checker to Stuyvesant, where Islam produced a document that appeared to be a Chase bank statement attesting to an eight-figure bank account. ...

But this is very ordinary fact checking.  Anybody can fake a bank statement.  The whole point of a bank reference is that you can verify it with the bank.  If New York Magazine had insisted on doing this they would likely have avoided this fiasco.

Sunday, December 7, 2014

Rape Fantasies

Last month Rolling Stone published a story about rape on college campuses which began with a lurid account of a University of Virginia coed being gang raped at a frat party.  Doubts were raised almost immediately about whether this account was accurate and as I write this Rolling Stone is not standing behind their reporting.  I didn't read the story before it was discredited but in hindsight it does read like something Stephen Glass (famous for inventing a whole series of stories for the New Republic catering to liberal biases) would have come up with.  Which means Rolling Stone should been wary and extra careful with their fact checking.  Which they do not appear to have been.

Some people find it hard to imagine a woman inventing a story like this.  I don't.  It is fairly common for men who have never been in combat (or in some cases even in the military) to invent stories about their horrific combat experiences (and the post traumatic stress which continues to affect them).   The motive of course being to attract the attention and/or sympathy that such stories naturally elicit (when believed).  I don't find it hard to believe that some women would invent rape stories for similar reasons.  And because rape accusations are fairly rare it doesn't take a large proportion of such women among the general population to make false rape accusations a significant fraction of all rape accusations.

As for Rolling Stone I suspect the problem is that among their staff it is generally believed that only "bad" people would ever question a woman's account of being raped.  This makes it hard for normal journalistic skepticism and fact checking procedures to operate properly.  If actually hiring a more ideologically diverse staff is too horrible for Rolling Stone to contemplate perhaps for each story they should designate a devil's advocate to bring up the objections that people are suppressing for fear of being thought sexist pigs (or whatever depending on the story).  This advice of course applies to any institution where there is a danger of legitimate questions being suppressed because of pressure to conform to a party line. 

Sunday, November 30, 2014

Goodell Management

NFL commissioner Roger Goodell's handling of the Ray Rice affair came under scrutiny again last Friday when an arbitrator reversed his indefinite suspension of Ray Rice from the NFL for punching his then girl friend (now wife). While Goodell's various decisions in this matter are certainly questionable the real problem is he shouldn't be making these decisions in the first place. There are two reasons for this. First it is not a good use of his time. Goodell's salary is about $40,000,000 a year. Assuming Goodell works 2000 hours a year this amounts to $20,000 per hour. It is unclear to me why the NFL owners think he is worth such a salary but it is surely not to personally conduct disciplinary hearings. The NFL could instead hire experienced respected professional arbitrators who could be expected to do a better job for a fraction of the cost (probably $1000 per hour or less). Second it exposes Goodell (and by extension the NFL) to needless bad publicity. These decisions will never please everybody but by assuming personal responsibility for them Goodell is providing critics with a ready target and focus for their anger. This is something that could and should be delegated and Goodell's failure to this reflects badly on his management ability.

Wednesday, November 26, 2014

Ferguson

As you may have heard a grand jury recently declined to bring charges against Ferguson police officer Darren Wilson who fatally shot Michael Brown on August 9, 2014. Most of the evidence the grand jury heard before reaching this decision has been released and is available here.  I have not examined this evidence in great detail but the case that Wilson acted criminally appears weak making the grand jury's decision unsurprising. 

Some people have doubted Wilson's account because Brown's actions (as related by Wilson) seem unlikely based on their experience of typical human behavior.  This objection basically makes no sense.  Getting shot by a cop is an unusual event and there is no reason to expect that the typical behavior of people about to get shot by a cop is similar to the typical behavior of people in general.  Particularly in this case as we know Brown was actually acting abnormally shortly before the shooting when he robbed the store. 

The prosecutor also acted abnormally by presenting the case to the grand jury without a recommendation.  It seems clear he did this because he didn't want to indict Wilson as if he had wanted to indict Wilson it is hard to see why he wouldn't have proceeded in the usual way.  The usual way being for the prosecutor to investigate the case, decide what charges he wants to bring and then present enough evidence to the grand jury to establish probable cause to bring these charges.  Probable cause basically means it that is more likely than not that the accused is guilty.  Since prosecutors generally don't want to bring charges that they don't think they can prove beyond a reasonable doubt at trial (and because grand jury decisions to indict don't have to be unanimous) it is rare for grand juries to reject the prosecutor's recommendation and refuse to indict.  By proceeding as he did the prosecutor risked having the grand jury unexpectedly bring charges leaving him in the awkward position of having to prosecute a case he didn't believe should have been brought but as a practical matter that was unlikely to happen.

I have seen suggestions that in cases like this the prosecutor should obtain an indictment whenever they can and let the trial jury decide the case.  In my view this is dangerous nonsense.  It seems obvious that it is completely unethical for a prosecutor to attempt to convict a person the prosecutor believes to be innocent.  So at least probable cause (in the prosecutor's view) is required.  Further in my opinion a prosecutor should not pursue cases in which they have reasonable doubt regarding guilt.  And of course a prosecutor has discretion to refuse to bring any case. 

That Wilson's actions were not criminal does not imply they were beyond reproach.  Opinions about politicized cases like this tend to become polarized leading to the "fallacy of the excluded middle" in which intermediate possibilities are ignored.  So for example a police shooting may be neither criminal nor a "good" shoot.  In my opinion criminal sanctions for police shootings are only appropriate for really egregious cases which this does not appear to be.   Attempting to impose criminal sanctions in borderline cases may even be counterproductive if it leads to police departments focusing on keeping their officers out of jail when police shootings occur rather than on reducing the number of police shootings.

According to this 2011 NYT article police shootings in New York City have dropped drastically in the last forty years.

The 33 instances in which an officer intentionally shot at a suspect last year represented a 30 percent decrease from the year before. But it reflected a far greater drop since the department began keeping these records in 1971, a year in which the police in New York City fatally shot 93 people and injured 221 others.

Last year, the police shot and killed 8 people and injured 16.

So it is possible to reduce police shooting without (so far as I know) sending a lot of officers to jail.  The article is based on the New York Police Department's (NYPD) annual firearms discharge report.  I suspect that just compiling this report annually by itself encouraged a reduction in police shootings.  (NYPD shootings have increased a bit since 2010 but remain way below the levels of 40 years ago).

In general mild sanctions widely imposed are a more effective means of altering behavior than harsh sanctions which are rarely imposed.  But of course harsh sanctions often have more political appeal.

Friday, November 21, 2014

The Money Culture

I recently reread another Michael Lewis book, his 1991 collection of magazine articles, "The Money Culture". While I generally like Lewis as a writer I didn't much care for this book. The pieces are dated and for the most part not all that great to begin with. Give this book a pass.

Tuesday, November 18, 2014

Procedure Codes

Back in March I went to the dentist for a routine cleaning. A week or two later I was reimbursed by my dental insurance (after my dentist submitted the claim directly). I noticed that as usual the insurance paid most but not all of the bill but otherwise paid little attention. I went back to the dentist in October for another routine cleaning. This time I paid a bit more attention when my reimbursement check arrived (because coincidentally dental insurance reimbursement rates had recently come up in conversation) and noticed it was more than the earlier check. My first thought was that my insurance company must have raised its reimbursement rates. This was a small part of the explanation but examining the respective explanation of benefits statements showed that most of the difference was because in March I had been reimbursed for procedure code D0190 (screening of a patient) whereas in October I was reimbursed for procedure code D0120 (periodic oral evaluation) which my insurance covers for a larger amount. (In both cases I also was reimbursed for procedure code D1110 (adult cleaning)). So my next thought was that my dentist had coded my claim suboptimally (from the point of view of maximizing insurance reimbursement) but my March bill showed the D0120 code and when I went into the dental office a couple of weeks ago with my bill and the insurance statement of benefits they insisted the insurance company would have received exactly what was on my bill. But they also offered to try and straighten things out. And Monday I duly received an additional reimbursement check from my insurance company.

So what happened? This is a bit puzzling as it seems everything should have been handled by computer with little room for human error given that the original entry into the system was done correctly as it appears it was. It seems unlikely that MetLife is randomly recoding procedures (to reduce reimbursements) and hoping nobody notices. I suppose a short lived computer software bug is the likeliest explanation but who knows. In any case when receiving explanations of medical benefits it seems it can pay to take a moment to compare the procedure codes to those on your bill from your provider.

Saturday, November 15, 2014

Flash Boys

As I mentioned in my previous post I recently read "Flash Boys" a 2014 book by Michael Lewis. Most of this book is about the creation of IEX, a recently formed dark pool (private stock exchange) which claims to be fairer to its clients than other stock trading forums.  A couple of chapters are devoted to the case of Sergey Aleynikov a computer programmer for Goldman Sachs who was criminally charged for allegedly stealing computer code after he quit to take another job.  His federal conviction was overturned on appeal but he still faces New York State charges.  These chapters have little to do with the rest of the book.

The founders of IEX have apparently constructed a narrative in which the previously existing US stock markets were rigged and corrupt inducing the outraged founders to create IEX to clean things up.  Lewis appears to have bought into this entirely and his book reflects this.  As I said in my review of Lewis's earlier book, "The Big Short":

There is a bit of a danger with an entertaining writer like Lewis that you will give undue credence to his point of view simply because he is such a good story teller. Lewis has been criticized along these lines but for the most part I think his books get the big picture (if not every detail) correct.

Lewis remains entertaining but in this book I don't believe he has gotten the big picture correct.  He doesn't seem to realize that if you try to buy or sell a large amount of a stock you will move the price against you (up if you trying to buy, down if you are trying to sell).  This is just how markets work by balancing supply and demand and doesn't mean they are corrupt or rigged.  But Lewis seems to think there is something nefarious about this process. 

On page 109 Lewis writes:

... After the market was computerized and decimalized, in 2000, spreads in the market had narrowed--that much was true.  Part of that narrowing would have happened anyway, with the automation of the stock market, which make it easier to trade stocks priced in decimals rather than in fractions. Part of that narrowing was an illusion: What appeared to be the spread was not actually the spread.  The minute you went to buy or sell at the stated market price, the price moved.  ...   

First decimalization narrowed spreads because as part of decimalization the tick size was reduced from .0625 (1/16) to .01 (not because decimal fractions have some inherent superiority to binary fractions).   It became apparent that the larger tick size had been keeping spreads artificially high and they narrowed when it became possible for them to do so.  Second the smaller tick size (and resulting narrower spreads) made it possible to observe stock prices to greater precision.  Which made it possible to observe smaller changes in stock prices.  In particular it became easier to observe the price changes induced by a moderate sized order which a larger tick size might have obscured.  This doesn't mean the narrower spread isn't real, it is quite real for small scale traders like most individual investors including me.

On page 114 Lewis writes:

... The algos had names like Ambush and Nighthawk and Raider and Dark Attack and Sumo.  Citi had one called Dagger, Deutsche Bank had Slicer and Credit Suisse had one named Guerilla, ... , Their very names made Rich Gates wary; he also didn't like how loudly the brokers selling them told him they'd come to protect him.  Protect him from what?  Why did he need protection? From whom did he need to be protected? ...

Although Lewis doesn't bother to explain it appears these programs mostly just disguise large orders by splitting them into smaller orders (enabling a better average price to be obtained).  As for from whom Gates needed to protected, apparently he needed to be protected from himself, from revealing his intent to buy or sell a large block of stock before he had to and thereby getting a worse price.

No doubt there are many ways US stock markets could be improved, the big players are all seeking private advantage.  But this unbalanced book while entertaining is not in my view a reliable guide to the issues.