The Volokh Conspiracy

My new post is up on the Reason Foundation web site. It’s about the Supreme Court’s decision last term in Filarsky v. Delia (decided almost exactly a year ago), which discussed whether a private attorney doing an internal-affairs investigation for a municipality was entitled to qualified immunity in an employee’s § 1983 lawsuit. Here’s an excerpt:

In 1871, as part of Reconstruction, Congress enacted a statute subjecting people to monetary liability if they violated anyone’s constitutional rights under color of state law. This statute, now called 42 U.S.C. § 1983, is a centerpiece of federal civil rights legislation, frequently used (especially since the 1960s) to secure compensation for victims of unconstitutional police tactics, mistreatment by corrections officers, bullying by heavy-handed bureaucrats, and the like.

But, despite its apparently absolute language, § 1983 hasn’t been interpreted to always require compensation. Rather, courts have read the statute as implicitly incorporating various forms of official immunity. For example, judges and prosecutors are absolutely immune from suit for any alleged violations of constitutional rights they commit as part of their official duties. The most important immunity is the “qualified immunity” enjoyed by executive officials. Officials with qualified immunity are only subject to suit if they violate a “clearly established right.” This is especially important when courts invent (or discover) new rights—or in areas like police or prison guard work, where the relevant constitutional standards (under, say, the Fourth or Eighth Amendments) are intensely fact-based, so courts often find violations that a reasonable officer wouldn’t necessarily have known about at the time. (All this also applies to Bivens actions, the more recent, judge-made, federal counterpart to § 1983 actions.)

Qualified immunity is a well-established part of civil rights law, though it remains controversial among scholars. On the one hand, a general rule that holds officials liable would better compensate victims, and may also lead to greater accountability. On the other hand, the fear of liability might make officials overly timid and might make it hard to recruit competent people for government work; moreover, courts might shy away from recognizing constitutional violations if they were concerned that doing so would excessively burden government functions.

Qualified immunity is usually given to public employees. But § 1983 is broader than that: private parties, for instance corrections officers at private prisons, can also act under color of state law, and thus can also be liable. Are these private parties entitled to claim qualified immunity? Yes they can, in many cases, said the Supreme Court on April 17, 2012 in Filarsky v. Delia. But reconciling Filarsky with previous decisions isn’t necessarily easy, and the availability of qualified immunity in the privatization context will probably continue to be confusing.

. . .

The Court’s language that Richardson didn’t apply to “the typical case of an individual hired by the government to assist in carrying out its work” (emphasis added) suggests that qualified immunity should be the norm for private individuals working for government, at least when they’re not organized in a corporate form and competing in a market. On the other hand, the Court’s opinion, as well as Justice Sotomayor’s concurrence, suggests that Richardson is still good law as to private prisons. Perhaps so. But, as Justice Sotomayor says, strong generalizations are elusive. What if Filarsky’s law firm had been hired in a competitive bidding process—would the case be like Richardson? What if a small municipality hired a local independent security guard as a temporary guard for its jail—would that be like Filarsky, despite a Richardson-like subject matter?

The chief judge of the Ninth Circuit, Alex Kozinski, who was on the 2003 panel that decided the now-reversed Gonzalez v. Spencer, recently said at an Emory Law Journal symposium on privatization that Filarsky puts Richardson on shaky ground. Nonetheless, Justice Breyer, who wrote Richardson, and Justice Ginsburg, who joined his opinion, joined the unanimous Filarsky decision, so at least two Justices (the other three in the Richardson majority have retired) think there’s a sound, even if fuzzy, distinction between CCA prison guards and occasional private lawyers working for municipal law enforcement. What’s certain is that between Richardson and Filarsky lies a substantial gray area, which will probably be the subject of substantial uncertainty and litigation in the years to come.

I’ve posted on the Reason Foundation web site on some previous occasions. You may be interested in these two:

• What a Recent Labor-Relations Decision Teaches Us About the Meaning of “Public” and “Private” — about an NLRB decision on charter schools, and different “public/private”-like standards under the NLRB, ADA, RLUIPA, and state action
• Supreme Court Antitrust Ruling Supports Public-Private Neutrality, Reduces Barriers to Privatization — about the Supreme Court’s recent decision in FTC v. Phoebe Putney, the antitrust state-action doctrine case

Click here to read the whole Filarsky v. Delia post.

As background, see my posts about Easter and Christian calendars (first, second, third, and fourth).

Recall that Easter is defined as the Sunday after the full moon after the spring equinox. The spring equinox could happen on March 19, 20, or 21, but for Easter-calculation purposes, we define the spring equinox as March 21. Why do we do this? These days, we can figure out exactly which day is the equinox, and everyone will come up with the same answer wherever they are on the Earth. But I’ll grant that, in past centuries, this would have been less precise and could have led (oh horror!) to wrongly placing the equinox too early, so if the full moon and Sunday happened immediately, Easter could have been celebrated before spring started. Moreover, it’s not just a matter of observing the sun in the current year, but also predicting when the equinox will fall next year and the year after that, so you can plan Easter years in advance. So perhaps defining the equinox as March 21 was justified as a way of building in a conservative assumption as to the start of spring, and producing uniformity from year to year.

But by artificially producing uniformity from year to year, the method produces disuniformity from place to place. If we went just by the equinox, people around the world would basically agree when that happened; at most, they would be off by a day or two. Same goes for the full moon. So most people would agree on Easter if you just observed the heavens rather than defining the equinox as March 21 (or using tables for full moons). But once you define the equinox as March 21, you enshrine the 13-day (and growing) difference between the Western and Eastern church calendars. So, ironically, what you adopted to produce uniformity is actually a major cause of the disuniformity in Easter observance that we see today.

This puts me in mind of a story that circulates among economists. It’s about an economist, I forget who, but probably someone who does contract theory. Contract theorists think contractual incompleteness is a big problem, and this guy knew how people building their houses often get into litigation with their building contractors over something or other that wasn’t clearly specified in the contract. He determined to avoid this, and write a very detailed contract that specified a lot more contingencies than the standard form contract. Surprise: he ended up in litigation with the contractor over some contractual incompleteness.

If that were all, it would just be a story about our inability to make everything complete. It doesn’t mean we can’t or shouldn’t try; conceivably he could have lowered his chance of litigation significantly, even if it didn’t help him this particular time. But actually it’s (apparently, from what I hear) more interesting. Most contracts are less incomplete than we think, because a lot of the incompletenesses are plugged by the common law of contracts. There’s a whole developed caselaw on what happens in particular situations that crop up a lot when the apparent contract falls short. I say “apparent contract” because the paper you sign isn’t the whole contract; the whole contract is the paper plus the whole body of law — defaults and the like — that fills the gaps.

It turns out that much of this common law of contracts was specifically designed around a particular standard-form contract. When the economist junked the standard-form contract and wrote a whole new one, he also (perhaps inadvertently) junked the common law that went with it. The result was that the gaps became a lot larger, and litigation more probable. The very act that was meant to reduce contractual incompleteness ended up increasing it.

There’s an analogy in here about Whack-a-Mole, or, in other words, spontaneous vs. planned orders, but that’s left as an exercise for the reader.

On Thursday, I posted about the difference between the Gregorian calendar (used by the Western churches) and the Julian calendar (used by most Eastern Orthodox churches). There’s a 13-day difference between them for the period between 1900 and 2100, so any given day shows up 13 days later on the Julian calendar than on the Gregorian calendar. So someone with a Julian calendar would celebrate October 12 on a day that shows up as October 25 on our Gregorian calendars.

Then, on Friday, I posted about the definition of Easter: the Sunday after the full moon after the spring equinox, and how that definition came to be. I brought up an initial puzzle: since the equinox and the full moon can be determined by observing astronomical phenomena, and we all have the same Sundays, why should Easter be any different for anyone? Sure, there are different calendars, but why should calendars be relevant for feasts that are defined in terms of astronomically objective things?

(Footnote: Yesterday’s post on “Western Christians, Orthodox Christians, and Jews: the coming convergence” was an April Fool’s joke, but contrary to what some might have thought, the previous posts weren’t just an elaborate setup — I didn’t even get the idea of doing an April Fool’s joke until the very morning of April 1.)

One answer was that “spring equinox” doesn’t mean spring equinox. The actual spring equinox could be on March 19, 20, or 21. But for Easter calculation purposes, we define the spring equinox to be March 21. This has the advantage that we don’t need to calculate every year, and Easter is guaranteed to be in spring (which apparently people thought was important). The disadvantage is that, by pinning the equinox to a calendar date rather than an astronomical phenomenon, you introduce a 13-day difference between the West and East based on the difference in their calendars. The Orthodox won’t consider that the Easter-calculation spring equinox happens until April 3 (even though they’ll have observed an actual spring equinox by then). So if a full moon happens to fall between March 21 and April 3, Western Christians will use that full moon to calculate Easter, while the Orthodox will use the next full moon, which introduces a difference of about a month (the lunar cycle is about 29.5 days).

Now we’ll talk about what “full moon” means. You won’t be surprised to learn that “full moon” doesn’t mean full moon. It’s an “ecclesiastical full moon,” which is defined as being 14 days after the “ecclesiastical new moon,” which is different than the actual new moon. Why do this? Well, you want to be able to figure out when Easter falls ahead of time rather than when you actually observe a new moon, so you want to be able to have tables to predict the new moon. Hopefully the tables correspond with astronomical reality, but the Council of Nicaea also thought it was important that all Christians celebrate Easter on the same date, and people in different places in the world might observe a new moon on slightly different days. So, in practice, we determine the ecclesiastical new moon by means of tables, and different tables (with different accuracies) are responsible for disagreements in calculating the date of Easter.

This stuff is so complicated that I’ll only describe the basic gist — there’s a big section on how to do the actual calculations in the Wikipedia article on the Computus. Consider a lunar month of 29.5 days. We could define a lunar year as consisting of 12 lunar months, with lengths alternating between 29 and 30 days. That gives us a year of 29.5 x 12 = 354 days. That’s 11 days shorter than a solar (non-leap) year.

So suppose that in 2006 we know the full moon falls on April 13. That means that, in 2007, there’s going to be a full moon on April 2, 11 days before April 13. And in 2008 (we’re ignoring leap years for this example), the full moon would be 11 days before April 2, which would be March 22. And in 2009, the full moon would be 11 days before that: March 11. That’s too early because it’s now before the spring equinox. So we add 30 days and get April 10. Then for 2010, we subtract 11 again and get March 30. And so on.

Then we introduce a ton of corrections. A correction for leap years (and the special Gregorian rule about leap years). A correction for the “adding 30 days” part from the previous paragraph, when a lunar month is only 29.5 days. A correction for the lunar month’s being slightly off from 29.5 days (it’s actually 29.53059 days). All this gets us a table predicting when ecclesiastical full moons (i.e., 14 days after ecclesiastical new moons) fall.

This works for most years. But comparing this Easter table with an actual table of full moons, we see that, for the period 2005-2013, it’s off in 2008 (Mar. 22 vs. Mar. 21), 2009 (Apr. 10 vs. Apr. 9), 2010 (Mar. 30 vs. Mar. 29), and 2012 (Apr. 7 vs. Apr. 6).

This particular Easter table is based on the Metonic cycle, named after Meton of Athens, who figured it out in the 5th century BC. This method is still how they figure out when to intercalate months in the Jewish (lunar) calendar. The monk Dionysius Exiguus published an adaptation of Metonic cycles into the Julian calendar in 525 (along the way he invented the practice of counting years [incorrectly] from the birth of Christ). His tables replaced earlier (less accurate) tables, for instance by Victorius of Aquitaine. It took a while for the Dionysian tables to replace the Victorian tables everywhere, and this difference is responsible for the differences in Easter calculation between the Irish church and the Roman church that were resolved at the Synod of Whitby in Northumbria in 664.

How do the Orthodox do it? They just use the Metonic cycle without any corrections, which departs from the full moon by more than three days every millennium. Now it’s a couple of days off, and it’s going to keep getting worse. Now some of the time, there’s no Sunday between the Western and Orthodox ecclesiastical full moons. So this doesn’t make any difference in terms of Easter. But the rest of the time, there is a Sunday between them, so Western Easter Sunday will be a week earlier than Orthodox Easter Sunday. Add that to the possible four-week error (actually, more than four weeks because of the 29.5-day lunar month), and you get a difference between Western and Orthodox Easters of 0, 1, 4, or 5 weeks.

Note, though, that the difference between the spring equinoxes will keep getting worse. And the difference between the full moons based on the Metonic cycles will also keep getting worse. You can see the results in Table 1 on this website, showing the differences between Western and Orthodox Easter. From 1583 to the 2000s, the differences are 0, 1, 4, or 5 weeks. After 2100, the four-week difference drops out, and we only have differences of 0, 1, or 5 weeks. By the 2400s, the differences will be 0, 1, 5, or 6 weeks. By the 2700s, the two Easters will never overlap, and we’ll have differences of 1, 2, 5, or 6 weeks.

Over the last few days, I’ve been blogging (here and here) about calendrical differences and how that affects the calculation of the date of Easter.

The good news is that these problems are going to diminish substantially pretty soon.

Recall the Orthodox calendar, which is 13 days off from the Western calendar. The Orthodox calendar is getting more and more off, by about 3 days every 400 years. But some of the Eastern Orthodox churches have adopted the Revised Julian calendar, also called the Milankovic calendar. The Revised Julian calendar omits 7 century leap years every 900 years, so there are 218 leap years instead of 225 every 900 years. (The two centuries that remain leap years are those that are equal to 200 or 600 mod 900, i.e., 2000, 2400, 2900, 3300, etc.) This makes an average year length of 365.24222 days. By contrast, the Gregorian calendar omits 3 century leap years every 400 years, so there are 97 leap years instead of 100 every 400 years. This makes an average year length of 365.2425; the two calendars will thus diverge by one day every 3600 years, on average. The true tropical year length is 365.24219, so the Revised Julian calendar is even a bit more accurate than our Gregorian one.

As I said earlier, it’s one thing to come up with a more accurate system; it’s another thing to get people to incur the transition costs. The Wikipedia article says that, so far, the Revised Julian calendar “has been adopted by the Orthodox churches of Constantinople, Alexandria, Antioch, Greece, Cyprus, Romania, Poland, and Bulgaria (the last in 1963), called the New calendarists. It has not been adopted by the Orthodox churches of Jerusalem, Russia, Serbia (including the uncanonical Macedonian Orthodox Church), Georgia, Mount Athos and the Greek Old Calendarists.” So there are basically three calendrical traditions out there: the Western Christian one (Gregorian), the traditional Orthodox one that’s in force in Russia (Julian), and the new Orthodox one that’s in force in Greece (Revised Julian), and there hasn’t seemed to be much movement toward convergence.

But it now looks as though more has been taking place behind the scenes than it seemed at first. John Paul II’s pontificate was marked by a rapprochement between Catholicism and Orthodox Christianity; for instance, the Pope apologized for the sack of Constantinople in 1204. Of course the Pope would have wanted the Orthodox to return to communion with Rome; that still doesn’t look like it’s going to happen, though individual Eastern churches, the so-called Eastern Catholics, have done so over the years: these include, for instance, the Armenian Catholic Church, the Melkite Greek-Catholic Church, and so on. But one area where there seems to be convergence is on the issue of calendars.

Seeing Catholics go back to the Julian calendar would be unthinkable — the Catholics invented the Gregorian calendar, after all, and it’s the Julian calendar that’s accumulating inaccuracy, so there’s no positive reason for adopting it. And seeing Julians go to the Gregorian calendar is unrealistic, because of the conservatism of Orthodox churches generally and the whiff of longstanding East-West animosity. On the other hand, the Revised Julian calendar has been a middle ground, culturally acceptable and at the same time astronomically more accurate.

So it was slightly surprising but not totally unexpected to see the new Pope issue, a bit in advance of Easter, the encyclical Quia pro unione, announcing the Vatican’s intention to switch to the Revised Julian system. This was already in the works under Benedict XVI, but the Vatican decided to wait for a new Pope just to reassure the other parties that the decision wouldn’t soon be reversed. (The buzz is that this decision came from Benedict XVI himself, who may have already decided at the time that he was on his way out.) At the same time, Patriarch Kirill of Moscow announced that the Russian church would likewise transition over. Both parties also announced their intention, in principle, to implement the command of the Council of Nicaea that all Christians celebrate Easter on the same day, but the details of that have yet to be hammered out. (I foresee either an adoption of the Dionysian tables or an actual reliance on lunar observation and computer-enhanced full-moon prediction.)

The Catholics won’t have to make any changes until 2800, which would be a leap year under the Gregorian calendar but not under the Revised Julian (which waits for 2900) — that’s almost 800 years. The Russians and others using the unreformed Julian calendar will make the 13-day jump “immediately” (in practice, it’ll probably be a few years, maybe up to a decade) but otherwise won’t have to change their practice until 2100. Until 2800, Catholics and Orthodox alike will be in synch with the secular systems, which all use the Gregorian calendar; after 2800, the difference will be one day for 100 years, then they’ll be in synch again for about 300 years, until 3200 — assuming the secular world doesn’t likewise change to Revised Julian.

What about Protestants? They’ll probably stay put, though perhaps the Anglican Communion will be next, and if that’s so, I would expect Lutherans to follow in the next century, and maybe everyone else within the next 800 years. Don’t count on the Jews to drop their lunisolar calendar anytime “soon,” though on the other hand, some prominent Reform and Conservative leaders have already put out feelers in that direction, citing as an advantage that Easter and Passover really ought to be synchronized.

UPDATE: Thanks to commenter arch1 for correcting 3100 to 3300.

Yesterday, I posted about the difference between the Gregorian calendar that we use in the West and the Julian calendar that’s still used by most Eastern Orthodox churches. There’s a 13-day difference between them for the period between 1900 and 2100, so any given day (e.g. October 12) shows up 13 days later on the Julian calendar than on the Gregorian calendar. So someone with a Julian calendar would celebrate October 12 on a day that shows up as October 25 on our Gregorian calendars.

Now, step 2 in understanding Easter: how is Easter defined? As an initial matter, note that Easter Sunday is the day when Jesus Christ rose from the dead, and the Gospels say that happened when he came to Jerusalem to celebrate Passover. So in principle, you could think that Easter happens at Passover — more precisely, on 14 Nisan, the day before Passover starts on 15 Nisan. (See also various verses in chapter 19 of the Gospel according to John.) Now the Jewish calendar is a complicated thing, but the bottom line is that 15 Nisan is always on a full moon after the spring equinox. How do they guarantee that this happens? Intercalation, that’s how. Because Jewish months are lunar, the folks in charge of the calendar stick in a whole nother month, called Adar II, before Nisan just to ensure that Passover is on a full moon after the equinox.

So why not tie Easter to Passover and celebrate it whenever 14 Nisan would fall in the Jewish calendar? The Christians who believed you should do this were called Quartodecimans, or “fourteeners” if you will. Various Christian communities followed the “14 Nisan” rule and just asked their local Jews when Passover started, but after controversies in the second and third centuries, Christians ended up settling the matter at the Council of Nicaea in 325.

The Nicene rule separated the Christian computation method from the Jewish one. In the first place, there was a concern over being in spring: though these days, 15 Nisan is always a full moon after the spring equinox, sometimes Easter ended up being celebrated before the equinox, apparently, said the bishop of Alexandria, “through negligence and error” on the part of the Jewish calendrical authorities. (It’s not necessarily negligence and error, because you could also choose to date the start of spring, as some Jews did, as the time the barley ripens. If you did that, Passover could sometimes fall before the equinox.)

In the second place, there was a concern over uniformity: apparently, not all Jewish communities at the time calculated their months in exactly the same way. This might actually be the same concern, if some communities used the equinox and others used the barley, or if barley ripened at different times in different places.

In the third place, reported Emperor Constantine: “It was . . . declared improper to follow the custom of the Jews in the celebration of this holy festival, because, their hands having been stained with crime, the minds of these wretched men are necessarily blinded.”

So Christian churches basically try and create their own “Christian Nisan” — figure out the full moon after the spring equinox — but also add on an extra rule, which is that Easter should fall on a Sunday. So that’s where we get the rule that Easter is the Sunday after the full moon after the spring equinox. (In principle, that still means that Western Easter should fall within Passover, but since Hillel II’s reforms in the 4th century, the Jewish calculations for Nisan are based on a formula and not on astronomy — see Gauss’s formula for the date of Pesach. Thus, in 2008, Western Easter fell on March 23, while Passover didn’t start until April 20.)

Given this formula, it’s not at all obvious why Christians should differ as to the date of Easter. We can observe the spring equinox using astronomy. We can observe the full moon using astronomy (though that might differ by a day or two depending on where one is in the world). And everyone has the same Sunday. So why should there be any (significant) difference?

The answer is that “Sunday” really does mean “Sunday,” but “full moon” doesn’t necessarily mean “full moon,” and “spring equinox” doesn’t necessarily mean “spring equinox.” The equinox rule is the biggest factor in the East-West date divergence. For purposes of calculating Easter, we use March 21 instead of the true date of the equinox, which could be March 19 or March 20. (The complexities behind “full moon” will be in a later post.) So we immediately see how the Western and Eastern churches can differ: March 21 is considered to fall on a different day depending on your calendar, and March 21 in the Julian calendar is what we in the West would call April 3.

Sometimes there’s no full moon between March 21 and April 3, so the relevant full moon for Easter-computation purposes is the same. For instance, in 2011, there were full moons on March 19 and April 18, so both calendars celebrated Easter on (Gregorian) April 24, the Sunday after (Gregorian) April 18. But sometimes there is a full moon between March 21 and April 3, so the relevant full moons will be about a month off. For instance, in 1997, the full moons were March 24 and April 27, so the Western churches celebrated Easter on March 30, the Sunday after March 24, while the Eastern churches celebrated Easter on April 27 itself (which happened to be a Sunday).

This explains an initial potential four-week difference between Western and Orthodox Easter. Stay tuned for why the difference could also be one week or five weeks.

UPDATE: Sorry, Christians out there, I accidentally wrote that Easter is when Jesus was crucified, not rose from the dead. Mea maxima culpa.

Easter is this Sunday, March 31. But Orthodox Easter, celebrated by most branches of the Eastern Orthodox Church (including the Russian Orthodox Church) and some Oriental Orthodox churches, is May 5 this year, a full five weeks after Easter as celebrated by Western Christian churches. (As a footnote, you may listen to Rimsky-Korsakov’s Russian Easter Overture at this site.) The five-week difference has happened recently in 2002, 2005, and 2008. But in 2001, 2004, 2007, 2010, and 2011, Western and Orthodox Easter fell on the same day. And in 2000, 2003, 2006, 2009, and 2012, Orthodox Easter was one week after Western Easter. What’s up with that?

First, let’s go over the basic difference between the Julian and Gregorian calendars. The Julian calendar, introduced in 45 BC, is easy: the year is 365 days long, except that we have a leap year every four years. After some confusion caused by off-by-one errors, apparently the proper sequence of leap years was reestablished by AD 4 or AD 8. (Good thing 4 and 8 happen to be divisible by 4: otherwise we might have leap years that occur on non-multiple-of-four years, which would be more confusing! Of course the numbering that established those years as being 4 and 8 came around much later, and was based on a totally different criterion.)

The Julian calendar had the advantage that it made the average year length 365.25 days rather than 365, which is closer to the astronomic truth. Without leap years, the years would be too short, so every four years the solstice would have a calendar day that seems a day later; instead of June 21, in 120 years we’d have it at July 21; and roughly 600 years after that, the summer solstice would be in December. All havoc would break loose. Dogs and cats would be living together. Also, it would make it hard to interpret old agricultural texts that tell you to do particular things on particular days.

Trouble is, the year isn’t quite 365.25 days long; the tropical year, which is the most relevant one for seasonal purposes, is more like 365.24219 days. So if you use the Julian calendar, your year is 0.00781 days too long, so after 400 Julian years, you’ll be 3.124 days off. Call it 3 days off, though let’s still remember, in the back of our heads, that it’s not exactly 3. So if a “true” calendar (where you’re constantly resetting the calendar to keep the solstices around the same days) and a Julian calendar were synchronized on January 1, 2000, then by the “true” date of January 1, 2400, the Julian calendar (with the longer years) would read something like December 29, 2399. The Julian guys would celebrate the New Year three days later than the “true” guys. Every 400 years, you get about 3 days more wrong, so the Julian dates seem to creep later and later relative to a “true” calendar. Before, the 365-day year had the problem that the date of the solstice was creeping later and later fairly rapidly (by a day every roughly 4 years); the Julian calendar overcorrects this slightly, so the date of the solstice creeps earlier and earlier fairly slowly (by, on average, a day every 128 years).

The Gregorian calendar fixes this problem by removing 3 days every 400 years. In practice, this is easy to implement: take the century years 2000, 2100, 2200, and 2300, and remove the leap days in 2100, 2200, and 2300. So of the century years, the only ones that remain leap years are 1600, 2000, 2400, etc., the ones divisible by 400. Now the average year length is 365.2425, which is off from 365.24219 by 0.00031 days, so the years are still too long, but by a much smaller margin, so the date of the solstice will creep earlier and earlier much more slowly (by, on average, a day every 3226 years). (See here for a more precise discussion of accuracy, which also discusses more technical points like the precession of the equinoxes.)

Now it’s one thing to figure out a calendar like the Gregorian one, which is more accurate and fairly easy to implement; but how do you actually get it adopted? The Gregorian reform was proposed by Pope Gregory XIII, and Catholic countries mostly adopted it in or shortly after 1582. By then, the error was 10 days; to prevent the backward creep of the date of astronomical phenomena, it was necessary to move the calendar date forward in one swell foop. The first places to adopt were Spain, Portugal, the Polish-Lithuanian Commonwealth, and most of Italy; there, September 4, 1582 was immediately followed by September 15, 1582. France adopted the calendar later in 1582, and other places adopted in 1583. The Protestant countries mostly adopted the new calendar over the course of the 18th century; England did it in 1752, by which time the error was 11 days.

The Orthodox countries of Eastern Europe lagged behind, though, and by 1918, the error was 13 days. In 1918, the new Soviet government adopted the Gregorian calendar; January 31, 1918 was followed by February 14, 1918. But the Russian Orthodox church didn’t change calendars, and, as the other Orthodox countries adopted the new calendar in the 1910s and 1920s, neither did their churches. (Some churches have adopted the Revised Julian calendar or Milankovic calendar, which is different than the Gregorian calendar, but is temporarily aligned with it.) So most Orthodox churches continue to follow the old calendar. Since 2000 was a leap year for both the Julian and Gregorian calendars, the error continues to be 13 days, and will continue to be 13 days until 2100. Thus, the date that the Orthodox churches consider to be December 25 is what we call January 7.

O.K., understanding this calendar issue is an important first step toward understanding the difference in the dates of Easter, but doesn’t quite get you there: after all, 13 days doesn’t sound much like the 0, 7, or 35 days’ difference that has prevailed in recent years. We’ll get closer to that in a subsequent post.

I have some questions for Jonathan (or anyone else), along the lines of my previous comments on DOMA and federalism:

1. Suppose DOMA defined “channels” to mean “navigable-in-fact waterways, as well as all interstate and intrastate waters including (but not limited to) intrastate lakes, rivers, streams (including intermittent streams), mudflats, sandflats, wetlands, sloughs, prairie potholes, wet meadows, playa lakes, natural ponds, drainage ditches, and ephemeral streams,” as Jonathan suggests. Jonathan suggests there would be something unconstitutional here because “the practical effects of this definition would be expand the scope of federal power so as to supplant state authority in an area of traditional state concern.”

I would think, though, that the problem isn’t with the definitional statute itself. For instance, suppose there were such a definitional statute, but it were never used anywhere — presumably no harm, no foul? Or what if it were only used in one statute somewhere in the U.S. Code, where Congress expressed its sense that channels were awesome things, or established a spending program to beautify channels. Again, nothing unconstitutional? I would think that any unconstitutionality would occur if and when that definition gets used in a way that regulates channels beyond what’s allowed under the Commerce Clause or any other source of federal power. So the problem isn’t with the definition but with the specific substantive provision where that definition gets used. There could be no such unconstitutional statutes, or one, or however many, but the fact that DOMA affects hundreds of statutes seems to play no role here.

Getting back to DOMA, I would then doubt that there’s any displacement of state authority as to marriage in any particular statute. Because, take the tax code: Congress didn’t have to grant special tax treatment, say, to married couples; Congress could have written the tax law so joint returns could be filed by any individual “and his blitiri,” where “blitiri” is defined to be “an opposite-sex member of the same union recognized by state law.” Or cut out the state law and imagine a statute with its own definition of “blitiri” cut out of whole cloth, like “a person of the opposite sex with whom the taxpayer has a romantic relationship” or “with whom the taxpayer cohabits” or whatever. Does the argument really hinge on whether the word in the federal statutes is called “marriage,” which happens to overlap with a word that’s used in state family-law statutes?

2. Or would the argument be different in these last two cases? Would the argument be that the tax statute would be unconstitutional on federalism grounds if the definition incorporates an arbitrary subset of state-law marriages, but not if the definition has no reference to state law? I would think that a definition that’s based on an arbitrary subset of state-law marriages (e.g. the limitation of qualifying marriage for immigration purposes to ones that are genuine in some sense, or a hypothetical limitation of some program’s benefits to beneficiaries’ spouses of more than ten years) is probably no less consistent with constitutional federalism than a definition that’s totally independent of state-law marriages, though quite frankly I’d think that in either case there are no constitutional federalism implications from the definition itself.

3. I think I agree that there’s no inherent power to define terms; the definition has to be necessary and proper for carrying out some power. Still, a definition in one statute that applies to hundreds of statutes can be supported by N&P hooked to a different specific federal power in each case, so the definitional statute itself isn’t supported by any one specific power. Suppose it’s true that Congress gave no thought as to whether the definition would make any of those hundreds of statutes work better. But suppose that, magically, the DOMA definition happens to be the most efficient one in each case. Then, I’d think, no harm, no foul, even if Congress put no thought into it. Am I right?

If so, then if there’s some violation of N&P based on the DOMA definition’s making the statute work worse, cost more money, etc., then again that must be because of some particular statute where this happens, not DOMA itself, so the fact that DOMA applies across the board likewise seems to play no role. There might be no such statutes, or one, or however many. Am I correctly interpreting the argument? If so, let’s challenge a particular statute as being beyond Congress’s powers based on the limitation of qualifying spouses, and bring in evidence of how the statute works so poorly that it exceeds N&P bounds, not DOMA itself.

4. Suppose that I’m right about points (1) and (3) and, magically, in all statutes where DOMA applies, i.e., in all statutes that use the term “marriage,” there’s no application that would run afoul of federalism if the DOMA definition had just been pasted in or if the term were called “blitiri,” and in all applications the DOMA definition is the most efficient of all possible definitions. (There are echoes of Leibniz here in more ways than one.) Would it make any difference that the purpose of the statute (either the subjective intent of the lawmakers, or the objective purpose as evidenced by the title of the statute) was to support traditional marriage, which (let’s say) isn’t within a listed power of Congress? I would think no, that would play no independent role. If the purpose of supporting traditional marriage were absent, and someone wrote an opposite-sex restriction into a single statute for some other reason, I’d think the constitutional analysis would go exactly the same, so if this statute were constitutional, DOMA would likewise be constitutional. Or does the argument put independent weight on the purpose? I’d think it shouldn’t.

As Jonathan wrote earlier, the Supreme Court’s Wednesday decision in Decker v. Northwest Environmental Defense Center might portend big things for Auer deference. Jonathan’s already hit the high points, but I wanted to note what Scalia said about how deferential Auer deference really is.

It turns out there’s disagreement in the literature over the strength of Auer (or Seminole Rock) deference, i.e., the deference that’s due to an agency’s construction of its own regulations, relative to Chevron deference, the deference that’s due to an agency’s construction of a statute (when the agency is making statements that have the force of law, as per Mead). Robert A. Anthony, in The Supreme Court and the APA: Sometimes They Just Don’t Get It, calls Auer “an indulgent if not downright abject standard of deference.” Bill Eskridge and Lauren Baer, in The Continuum of Deference: Supreme Court Treatment of Agency Statutory Interpretations from Chevron to Hamdan, classify Seminole Rock/Auer deference as stronger than Chevron deference. Stephen M. Johnson, in Bringing Deference Back (But for How Long?): Justice Alito, Chevron, Auer, and Chenery in the Supreme Court’s 2006 Term, also describes Auer deference as “even more deferential than Chevron deference.”

But not everyone takes this view. Richard Pierce, the author of one of the leading administrative law treatises, writes in Democratizing the Administrative State that “Seminole Rock deference is about as strong as Chevron deference.” (All this is taken from Hanah Metchis Volokh, The Anti-Parroting Canon, 6 NYU J. L. & Liberty 1 (2011).)

Scalia, in Decker, comes down on the Pierce side:

The canonical formulation of Auer deference is that we will enforce an agency’s interpretation of its own rules unless that interpretation is “plainly erroneous or inconsistent with the regulation.” But of course whenever the agency’s interpretation of the regulation is different from the fairest reading, it is in that sense “inconsistent” with the regulation. Obviously, that is not enough, or there would be nothing for Auer to do. In practice, Auer deference is Chevron deference applied to regulations rather than statutes. The agency’s interpretation will be accepted if, though not the fairest reading of the regulation, it is a plausible reading—within the scope of the ambiguity that the regulation contains.

Of course, Scalia is opposed to Mead and the whole idea of different levels of deference (Chevron, Skidmore, etc.). But even if you accept Mead, you could still think that you can never be more deferential than Chevron. If there’s an ambiguity, Chevron allows for interpretations that are within the ambiguity, but surely you can never allow for interpretations that are outside of the ambiguity. So, on Scalia’s view, Auer and Chevron are equally strong.

My new post is up on the Reason Foundation web site. (Earlier, I had posted there on the Supreme Court’s antitrust decision in FTC v. Phoebe Putney.) Here’s the beginning:

In June 2010, the Chicago Alliance of Charter Teachers & Staff filed a petition to be certified as the exclusive bargaining representative of teachers at the Chicago Mathematics & Science Academy (CMSA), a charter school. The petition was filed with a specialized state agency—the Illinois Educational Labor Relations Board (IELRB)—presumably because that’s the agency that enforces the Illinois Educational Labor Relations Act (IELRA), which, by its terms, extends to “educational employer[s],” a term defined to include charter schools as well as public schools.

The charter school, though, thought otherwise. The school’s position was that it was subject to federal, not state, labor law—specifically, that it was governed by the National Labor Relations Act (NLRA), enforced by the National Labor Relations Board (NLRB). The NLRA, after all, extends to all “employer[s]” with a handful of exceptions. The only arguably relevant exception is the NLRA doesn’t cover “a political subdivision” of a state; but the charter school argued that, unlike a public school, it isn’t a political subdivision, so the only labor law that applies is the federal NLRA. Because of federal supremacy, anything in the IELRA to the contrary is preempted, and IELRB jurisdiction is inappropriate.

Why should a school care whether it’s under the jurisdiction of a state or a federal agency? Federal and state labor law differ in various ways, but one relevant consideration is possibly that federal law requires secret ballots when an employer doesn’t recognize the union voluntarily, while state law also allows for union recognition using the non-secret “card-check” method. And employers are far more enamored of secret ballots than are unions. In any event, on December 14, 2012—after initially losing before the NLRB’s Acting Regional Director—the school won before the full NLRB. CMSA isn’t a political subdivision of a state, the NLRB held, so the school is indeed subject to the NLRA.

To decide this, the NLRB applied its longstanding test that “an entity may be considered a political subdivision if it is either (1) created directly by the state so as to constitute a department or administrative arm of the government, or (2) administered by individuals who are responsible to public officials or to the general electorate.” As far back as 1967, the NLRB applied this test to conclude that a local natural gas utility in Tennessee wasn’t a political subdivision because (1) it was privately incorporated and (2) its officials were neither state-appointed nor elected. The Supreme Court decided in 1971 that the NLRB had misapplied its own test and that the natural gas utility was a political subdivision: on prong 2, though the officials weren’t appointed by the state, they could be removed by the governor, which made them “responsible to public officials.” But the Supreme Court didn’t question the correctness or reasonableness of the NLRB’s test itself, and the NLRB has continued to apply the test in the four decades since.

Applying this test to the Chicago charter school, the NLRB held that, though the state of Illinois created the category of “charter schools” and chooses whether to charter any particular school, the school itself is incorporated by private individuals under the state nonprofit law and exists prior to the grant of the charter. Thus, on prong 1, the school wasn’t created directly by the state. On prong 2, the NLRB held that being “responsible to public officials” means being “appointed by or subject to removal by public officials.” (Note how the addition of “removal” fixes the mistake identified by the Supreme Court in 1971.) And while charter schools are regulated by the state according to the terms of their contract (the charter) and the charter school law, the school is run by a board whose members are appointed and removed purely privately. The school, the NLRB held, is thus an “employer” under the terms of the NLRA and is thus subject to federal, not state, labor law. (One dissenting board member suggested that the NLRB should nonetheless have used its discretion chosen to decline jurisdiction, leaving the matter to the state of Illinois for the sake of federal-state comity; but the NLRB majority decided not to pursue this course.)

The rest of the post goes on to discuss the ADA, RLUIPA, and the state-action doctrine. My conclusion:

If this is confusing, that was the goal. Contracting with a private corporation to deliver an activity may alter the labor-management relations regime and the ADA regime but not the RLUIPA regime. The constitutional state-action regime may be unchanged with respect to the population served, but may be radically different with respect to hiring and firing decisions. What other regimes are altered may also depend on what conditions the contractor committed to fulfill in its contract. How, then, do we define “public” vs. “private,” “government” vs. “private sector,” “instrumentality of government” vs. “mere contractor”, “state actor” vs. “private actor”? We don’t. Each of these terms is a shorthand designating a broad set of attributes, and contracting out is all about exploring the limits of these concepts. For any given statute or constitutional provision, different rules might be appropriate in different contexts. Choosing to contract with private corporations may change the ownership of assets, but other legal regimes may or may not change, and overall accountability might either increase or decrease. From the perspective of the institutional designer, better just to ask how institutions—whether schools, prisons, or anything else—should be run, and choose the set of governance regimes appropriate to the goal.

Read the whole thing here.

UPDATE: Corrected a misstatement of labor law, thanks to commenter JFT.

My friend Marvin Ammori has a piece in Slate about Uber, Lyft, and SideCar, three new companies trying to provide competition to the taxicab market. Here’s what he has to say about the antitrust solution to taxicab monopolies:

It might seem odd to call for a federal agency like the FTC to take action in what appears to be a purely local issue. But not only does the FTC have the authority to take these cities to impartial federal courts and end their anticompetitive actions; it also has deep expertise in taxi markets and antitrust doctrines. By law, the FTC’s power to regulate “interstate commerce” is just as broad as Congress’, and that power is famously far-reaching, even covering the growing of plants at home for purely personal use. Here, as the agency noted in its 1984 cases against Minneapolis and New Orleans, the FTC could regulate local taxi markets merely because interstate travelers take taxis (and Ubers) to and from the airport when they fly across state lines. Moreover, Uber, SideCar, and Lyft are California-based technology companies competing in multiple states.

The FTC’s general antitrust authority, however, only sometimes outranks city authority. Under the Constitution, federal law trumps both state and city law. But antitrust law allows states some exceptional leeway to adopt anticompetitive business regulations, out of respect for states’ rights to regulate business. This federal respect for states’ rights does not extend to cities—they are subject to the antitrust laws unless states pass legislation absolving them. For example, Minneapolis ended up avoiding the 1984 lawsuit by acceding to the FTC and permitting more competition. New Orleans did something different: It succeeded in lobbying the state of Louisiana to authorize its anticompetitive actions, to the detriment of consumers.

Some states, including Florida, Nevada, and Texas, could argue that like Louisiana, their laws currently authorize the city governments to undermine competition in their taxi markets. Nonetheless, the FTC should test the law in these states with litigation, as the Supreme Court’s position is unsettled. Under the few Supreme Court decisions spanning the past several decades, cities can engage in anticompetitive acts only when the state legislature “clearly articulated” and “affirmatively expressed” a state policy to displace competition and replace it with an uncompetitive market. To complicate matters, the court “disfavors” interpretations of state law that permit anticompetitive city regulation, yet it still permits city anticompetitive action even it is merely a “foreseeable result” of the state law. Put all this together and the FTC can likely argue that state taxi authorizations do not foresee anticompetitive restrictions on apps like Uber and services that permit average citizens to tip one another for a ride. At the very least, Washington, D.C., is not in a state, so no state law can protect it from the FTC’s antitrust powers.

Since the FTC/antitrust solution may not work, Marvin suggests action by state governments or Congress, which he argues “should be politically feasible.”

Yesterday and today, I attended two excellent talks in the Atlanta area — the free-market/libertarian scene in Atlanta is actually kind of booming these days, with FEE, AFF, GPPF and ATLOS having an active presence, as well as some interesting student organizations.

Last night, Tom Palmer of Cato/Atlas came down to Kennesaw State University to talk about his new edited volume, After the Welfare State. I met Tom 20 years ago at an IHS conference, and he’s always an excellent speaker. (You can get the book from Amazon, but Students for Liberty and others are also giving it out at events, to student groups, etc.)

Then, this morning, Yaron Brook of ARI gave a talk on “Morality and the Marketplace” as part of his book tour for his (and Don Watkins’s) new book, Free Market Revolution: How Ayn Rand’s Ideas Can End Big Government. I’d never met Yaron or heard him speak before, and (not being an Objectivist) I wasn’t sure whether I’d like it, but it was an outstanding speech, from which one may possibly extrapolate that those who take an interest in this sort of thing might enjoy his book. The Georgia Public Policy Foundation’s monthly policy breakfasts and other events have been consistently interesting, from what I’ve seen of them so far.

I just finished reading Michael Chabon’s The Yiddish Policemen’s Union, which I loved. It turns out that both this and the last English novel I read, David Bezmozgis’s The Free World, both have Esperanto components to the plot, which was unexpected but not totally out of left field, since both books have to do, broadly speaking, with the intellectual interests of twentieth-century Jews.

Is it just me, or did anyone else immediately imagine the Bina Gelbfish character as Alex Kingston, aka River Song, the Doctor’s Wife? [UPDATE: Commenter MathAdder points out that The Doctor's Wife could refer to this, but see also this.]

Soothsayer. Caesar!

CAESAR. Ha! who calls?

CASCA. Bid every noise be still: peace yet again!

CAESAR. Who is it in the press that calls on me?
I hear a tongue, shriller than all the music,
Cry ‘Caesar!’ Speak; Caesar is turn’d to hear.

Soothsayer. Beware the ides of March.

CAESAR. What man is that?

BRUTUS. A soothsayer bids you beware the ides of March.

CAESAR. Set him before me; let me see his face.

CASSIUS. Fellow, come from the throng; look upon Caesar.

CAESAR. What say’st thou to me now? speak once again.

Soothsayer. Beware the ides of March.

CAESAR. He is a dreamer; let us leave him: pass.

In honor of Pi Day (3/14), I’ll share a mnemonic for 167 digits of pi, which I developed with some friends in the mid-’90s.

Many people know the common mnemonic: “How I need a drink, alcoholic of course, after the tough lectures involving quantum mechanics.” Count the number of letters in each word and stick a decimal point after the 3, and you get: 3.14159265358979, which is already more than you need for most purposes. The thirty-second digit after the decimal point is a zero, which in this mnemonic we represent by the end of a sentence. Here’s 167 digits’ worth:

How I need a drink, alcoholic of course, after the tough lectures involving quantum mechanics; but we did estimate some digits by making very bad, not accurate, but so greatly efficient tools! In quaintly valuable ways, a dedicated student — I, Volokh, Alexander — can determine beautiful and curious stuff, O! Smart, gorgeous me! Descartes himself knew wonderful ways that could ascertain it too! Revered, glorious — a wicked dude! Behold an unending number: pi! Thinkers’ ceaseless agonizing produces little, if anything! For this constant, it stops not — just as e, I suppose. Vainly, ancient geometers computed it — a task undoable. Legendre, Adrien Marie: ‘I say pi rational is not!’ Adrien proved this theorem. Therefore, all doubters have made errors. (Everybody that’s Greek.) Today, counting is as bad a problem as years ago, maybe centuries even. Moreover, I do consider that variable x, y, z, wouldn’t much avail. Is constant like i? No, buffoon!

Note that the word “greatly” in the first sentence was originally “f**king”, but this is a family blog.

This was written up in the March 18, 1996 issue of The Scientist magazine (at the link you’ll also find a picture of me from back then) and was also mentioned in Ivars Peterson’s MathTrek (possibly called Mathland then) from March 11, 1996. As a stocking stuffer, you might try David Blatner’s The Joy of Pi (1997), which quotes me on the subject.

Antreas Hatzipolakis compiled a list of pi mnemonics in the late 1990s, and some other pi links are here. One of the best pi mnemonics, which gets up to 740 digits, is a retelling of Poe’s The Raven; here’s its first stanza: [UPDATE: Commenter raederle points out that this is actually the beginning of a super-long work called Cadaeic Cadenza by Mike Keith, which is itself a mnemonic for pi and seems to have about 4000-5000 words.]

Poe, E.
Near A Raven

Midnights so dreary, tired and weary.
Silently pondering volumes extolling all by-now obsolete lore.
During my rather long nap – the weirdest tap!
An ominous vibrating sound disturbing my chamber’s antedoor.
“This”, I whispered quietly, “I ignore”.

Of course, a classic, even if it doesn’t cite me, is Petr Beckmann’s A History of Pi (1976).

White smoke. Live video feed here. I’m predicting Angelo Scola, archbishop of Milan.

UPDATE: Jorge Mario Bergoglio of Argentina, who will be Francis? I guess he doesn’t have a number, then.

UPDATE 2: Here’s Pope Francis’s Wikipedia entry.

UPDATE 3: Man, Google News and the N.Y. Times don’t even have the pope’s name yet. Advantage: Volokh Conspiracy! Also, Wikipedia.