No Panaceas

Friday, March 28, 2003
Oyezology: I’m going to start picking on reporters a bit by pitting their oral argument predictions against the actual outcomes. As I talked about the other day, reporters often like to use the oral arguments to make predictions about Supreme Court case outcomes. As I discussed, to me this is a rather specious exercise. (Though admittedly it is something that is rather hard to resist doing after hearing a spirited argument.) On occasion I plan to track a few examples of reporter’s making predictions on major cases, and then check back later to see the result. Who knows, maybe we’ll find out that their predictions aren’t so bad.

I’ll start with Lawrence v. Texas, the case heard Wednesday. I’m regretting not camping out to hear the arguments because it sounded like a rip-roaring good time. Well, we are talking about the Supreme Court here, so maybe rip-roaring is a stretch. Regardless of what you think of her politics, Dahlia Lithwick at Slate always provides the most entertaining summaries of Court arguments. Here’s her piece on Lawrence v. Texas.

I don’t know what to think about the importance of oral arguments. Do they really make a difference? That is, is the outcome of a case affected by a particularly eloquent, or particularly ineloquent or incompetent, argument? I don’t know, and I am not aware of any systematic research on the question. Indeed, I’m not sure how one could design a study to test the idea. I am sure there are anecdotes that show both sides. In Storm Center David O’Brien seems to conclude that oral arguments do matter. He stresses a quote by Justice Anthony Kennedy that’s worth repeating, “Does oral argument make a difference? Of course it makes a difference. That’s the passion and the power, and the poetry of the law – that a rhetorical case can make a difference, because abstract principles have to be applied in a real-life situation. And that’s what the lawyer is there to remind the Court about” (Source: O’Brien, David. 2000. Storm Center. New York: Norton, p. 261).

Regardless of whether or not it affects the case’s outcome, I agree with a lot of observer’s contention that Charles Rosenthal, arguing for Texas, did an abysmal job Wednesday. It is stunning how often lawyers appear before the Court utterly unprepared to answer the most basic and predictable questions. Why does this happen? My guess is that – oh shock – it has a lot to with politics. Here’s an assertion: The worst arguments tend to come from elected state or local officials. (Rosenthal is the District Attorney from Harris County (Houston), Texas.) Politicians want the exposure and resume enhancement that arguing before the Court brings. Ego and myopia blinds them to the fact that dog-and-pony shows that play well before juries and electoral audiences doesn’t play so well before nine of the smartest people in the country. I have trouble believing that Rosenthal was the best the Harris County DA's office, or the state Attorney General's office, had to offer. Just a thought.

Anyway, unfortunately almost everyone I’ve found thus far is saying that the orals indicate that Lawrence v. Texas is a toss-up. The most important exception is Linda Greenhouse of the New York Times: “A majority of the Supreme Court appeared ready today to overturn a Texas "homosexual conduct" law…” I’ll post any other clear predictions that I come across.

Wednesday, March 26, 2003
They are sold out down at the Marble Temple: It's getting hard these days to get in to hear the Supremes. I went down bright and early this morning to see arguments for Texas v. Lawrence. No chance. Press and VIP passes left only a handful of public seats to a hardy group of overnight campers. The Michigan cases next week will be even worse.

It's too bad. I was looking forward to a spirited rendition of the "Penumbras of Privacy Polka," a piece that's really fallen from the repertoire since Rehnquist became the host. And then there's the always popular gameshow skit: "Who Wants Equal Protection?"

In big cases like this the media always tries to prognosticate the outcome based on the justices' behavior during the arguments. This is a bit like the way Kremlinologists used to interpret internal Soviet politics based on things like who was standing next to whom on the Kremlin wall during parades. It's rarely very useful. Sometimes individual justices are playing devil's advocate, trying out ideas, or just generally picking on a hapless attorney. Plus you can't really interpret a justice's silence as indicative of anything particular. Maybe I'll add a feature here where I describe individual reporter's tea-leaf reading and then come back after the decision and see how the fortune turned out.

Speaking of the Court. Here's a highly informative blog I just came across.

Tuesday, March 25, 2003
Please forgive this parochial moment: The Rice Owls are number one in all the polls and the ISRs to boot.

Monday, March 24, 2003
Moving on Up: The rally effect is underway. Gallup has Bush's approval rating at 71.

Approval Redux: The Post ran an article yesterday on the impact of the war on Bush's approval. See my posts of 3/20 and 3/21 for my take on this. The article's really rather silly. The current war would not have a 9-11 style long-term impact on Bush's approval even if it were less controversial.

March Lameness II: Last Monday I talked about March Madness in light of the intransitivity problem. The ultimate winner will be champion because it is a good team but also because of favorable matchups. Excepting the unlikely existence of a "Condorcet winner," different bracket arrangements would yield different champions. As a consequence, the "champion" who is crowned in a couple of weeks will be mythical in no less a sense than, say, a team that is voted champion in lieu of a tournament.

Actually there’s another big problem with March Madness. I’ll call it the small-n problem. Suppose you flipped a coin and it came up heads. Would you then conclude the coin was biased in favor of heads? No. So why are we supposed to conclude that Butler, say, is better than Louisville simply because Butler beat Louisville in a single game? Yet, implicitly that is what we are doing here. We are assuming that these single-elimination matchups identify the better teams. The problem is that there is some amount of randomness in all sports. There is, for example, a certain amount of (hopefully random) error in official’s calls. This randomness may make the difference in a single game. Over many games, however, the randomness will, well, play out randomly without any particular advantage to either team.

Consider two teams, one clearly better than the other. Let’s characterize the teams by probabilities of winning a single game against the other team. Let’s say that the stronger team has a .60 probability of winning a game against the team. (Thus the weaker team has a .40 chance of winning.) What does this mean? Well, for starters it suggests that we would predict the stronger team to win the single game, just like weather forecasters would predict rain for a day that has a .60 probability of rain. Yet, as with weather forecasts, the prediction will often prove wrong. In fact, we would expect the weaker team to win four out of ten matchups. (Already this is starting to sound like March Madness.) That is, four out of ten times the lesser team wins the game. This means that four out of ten times, with some variance, a single elimination tournament “chooses” the wrong team to advance to the next round.

However, if the two teams played ten times, then we would expect the stronger team to win six of the games. This is a prediction; randomness will sometimes prove the prediction incorrect. For example, if you flip a coin ten times it might come up five heads, five tails. But it might also come up six heads, four tails. It might even occasionally come up seven heads, three tails, or even eight heads, two tails. There is going to be some variance even with a fair coin.

So how many games do two teams have to play before we have a pretty good proof that one is better than the other? A computer simulation can give us a precise answer. Using a binomial distribution I simulated three-game, seven-game, twenty-five game, and ninety-nine game series between the .60 and .40 teams discussed above. For good measure I then did it with teams with respective .75 and .25 chances of winning. I ran each series 1000 times to see how often the weaker team won. (This is easier to do than it sounds.) Here are the results:

As you would expect the chances of a “wrong” winner go down with the more games played. A team with a .40 chance of winning any one game will win more than one-third of the best-of-three series played, and more than one-quarter of the best-of-seven series played. The numbers are notably smaller for the .25 team, but even here the wrong winner takes the best-of-three series one out of every six times or so. This is why stronger teams do not like to play best-of-three series, and it is part of the reason why the NBA just scrapped the best-of-three first round. (Though I am sure the main reason is "more games = more money.")

So what am I really saying? Am I suggesting that we need to hold best-of-ninety-nine game series? Obviously not. A really long series would suck the life out of the game and, besides, the intransitivity problem would remain anyway. I think seven game series are fine for something like the NBA. (Though it is actually a little short for baseball, where the pitching issue comes into play.)

For college basketball a tournament of seven game series is not practical. And besides, March Madness is exciting exactly because the single-elimination structure produces so many upsets. But, and this is the point, there has long been this conceit that tournaments are a superior way to choose champions, and that college football should follow in the footsteps of college basketball and institute a tournament. That’s nonsense. The champion that emerges April 7th will be just as mythical as an elected champion.