TO: Commission on Structural Alternatives for the Federal
Courts of Appeals
FROM: Andrew J. Kleinfeld, Ninth Circuit Judge
RE: Splitting the Ninth Circuit
DATE: May 22, 1998
I suggest that the Ninth Circuit be divided as the Murkowski bill would, placing
Alaska, Washington, Oregon, Idaho and Montana in a new Twelfth Circuit. This would leave
the Ninth Circuit very large, as any circuit that includes California must be. The significant
problems of size would be solved, however, for five other states.
Our existing Ninth Circuit has many of the best appellate judges in the United States.
We have had a succession of superb chief judges. In my appellate experience as a practicing
lawyer, I never saw panels of more capable people who prepared so well as our court. Yet our
reputation does not reflect this high quality. I was recently sent a clipping from an Owensboro,
Kentucky newspaper characterizing us with phrases (some quoted from the Supreme Court)
such as "sharply rebuked," "grave abuse of discretion," "stinging lecture," "scolded the lower
court," "procedural legerdemain," and "negligence." Despite the high quality of our judges,
our Supreme Court reversal rate is notorious, 28 of 29 cases reviewed last year. And that does
not include cases where the Court affirmed another circuit and noted that our conflicting view
Why are judges so good making so many errors? That is a serious question, and I do
not think, in view of our Supreme Court record, that it can be avoided. The reason, I suggest,
is our size. We have so many judges that we cannot read each others' opinions, and we cannot
correct errors by effectively rehearing cases en banc.
In every organization, there has to be some size limit, at the bottom end and the top
end, such that if the organization was smaller or bigger, it would be less efficient. That is
probably why dentists' offices do not grow to the size of General Motors, and General Motors
is not operated out of a small garage. Justice Brandeis wrote, in The Curse of Bigness, "In
every business concern there must be a size limit of greatest efficiency. What the limit is will
differ in different businesses and under varying conditions in the same business. But whatever
the business or organization, there is a point where it would become too large for efficiency
and economic management, just as there is a point where it would be too small to be an
efficient instrument. The limit of efficient size occurs when the disadvantages attendant on the
size outweigh the advantages, and for large size, when the centrifugal force exceeds the
Economists often conceptualize the difficulties of size as information costs and agency
costs. These costs limit economies of scale that may be obtained by larger size. The concepts
are helpful in analyzing large courts of appeal. Information costs are those costs attendant
upon gathering the necessary information on which to act. The relevant information cost for
our purposes is the time it takes for appellate judges to learn what the law of their own court
is. Agency costs are the costs of getting those who serve the organization to carry out their
tasks correctly. The relevant agency costs are the information cost of finding out what other
judges are writing, and the additional time it takes to bring mistaken decisions into compliance
with the law.
Reading each others' decisions.
Judges on the same court should read each others' decisions. We are so big that we
cannot and do not. That has the practical effect that we do not know what judges on other
panels are deciding. It is odd word usage to call a public body a "court," in the singular, if its
judges do not ever sit together as one body, and do not even read each others' opinions.
The Ninth Circuit issues too many decisions for anyone to read. We learn of decisions
of our own court, affecting our own cases, when they are cited to us by lawyers or by our law
clerks. This deprives those we serve of two important values a court should provide. We lack
continuity in our understanding of our own court's views. And we lack opportunity to obtain
correction of mistakes in other panels' decisions, by internal suggestions or en banc procedures.
The Ninth Circuit terminated 4,841 cases on the merits over 12 months, according to
the 1997 Annual Report of the Director of the Administrative Office. Assuming 200 work
days available for reading decisions (most of us cannot do routine reading while on calendar,
and do not work seven days out of every week of the year), that would require each of us to
read 24 decisions a day, in order to keep up. That is impossible, if enough time is given to
each to understand it and spot errors, and if any other work is to be done. If we ignore the
unpublished decisions (as most of us are forced to do, allowing for much error to go
uncorrected in them), there were still 872 published dispositions, each with precedential force.
Keeping up would require us to read a little over four per day, manageable if one is not on
calendar, but generating a pile of about 22 plus the new ones that come in on Monday after a
week on calendar. At that point, the opinions can only be glanced at to see if they affect
pending cases or resolve matters in which the judge happens to have a particularly strong
Note that this problem of reading opinions is purely a function of size. It has nothing
to do with the rate at which judges produce opinions. If each judge generates 32 published
decisions per year, that produces a reading load of a little over one per day on a seven judge
court, and about four and a half on a 28 judge court. Even if everything else about a large
court and a small one is the same, the judges on the court can read their own court's decisions
only if it is small. I use seven as my hypothetical size because that is the proposed size of the
Twelfth Circuit in the Murkowski bill.
This high cost of information because of size affects not only us, but also district
courts, practicing lawyers, and, most important, the general public. The great scholar of the
common law tradition, Karl Llewellyn, characterized the chief virtue of appellate opinions as
providing "reckonability of result." Professor Llewellyn said "spend a single thoughtful
weekend with a couple of recent volumes of reports from your own supreme court, . . . and
you can never again, with fervor or despair, make that remark about never knowing where an
appellate court will hang its hat. Spend five such weekends, and you will be getting a
workable idea of the local geography of hat racks." Karl N. Llewellyn, The Common Law
Tradition: Deciding Appeals, 179 (1960).
When I was a practicing lawyer, I was generally able to predict with great accuracy
what the Supreme Court of my state would do, even where there was no case in point. I read
its decisions as they came down, and as Llewellyn suggests, I knew what the thinking process
of each of the five justices would be when faced with a new problem. All the justices sat on
every case, as is typical of state supreme courts, so there were no unpredictabilities generated
by not knowing who would be on a panel. Of course, there were occasional surprises, but not
very many, to me or to other lawyers. Our clients benefited from advice based on this high
degree of reckonability. They could learn from their lawyers what the law was, and did not
have to spend money or go through the misery of litigation to find out.
The Ninth Circuit, because of its size, is not and cannot be a reckonable court. No
district judge and no lawyer can, by reading even a few hundred of our decisions, predict what
our court will do in the next case. Even if the decisions could be read, there are over 3,000
combinations of judges who may wind up on panels, so the exercise would not be worth the
time. At best, the bar can predict that we will restate our clear holdings as controlling law,
though different panels may apply the same holdings to similar facts in different ways. The
disparateness will naturally be higher in unpublished dispositions.
A court that is not reckonable is of far less use to the general public, the lawyers who
represent them, and the trial judges who must adjudicate their cases, and far more expensive,
than one that can be predicted. People sometimes talk of the expense of the judiciary as
salaries, buildings, air fares, and so forth. Those are a tiny fraction of the expenses occasioned
by cases in courts. The expense of counsel to litigate, and the expense of uncertainty and time
for the parties engaged in litigation, is far greater than the out of pocket expense of the court
system to the taxpayers. There is no expense caused by the law that can be so great as the
expense of unpredictability that requires litigation.
Nor is this problem a new one. Speaking of his days on the old Fifth Circuit before it
split, Judge Tjoflat explained that "`If you have three judges on a court of appeals, the law is
stable. . . . When you add the fourth judge to that court, you add some instability to the rule of
law in that circuit because another point of view is added to the decision making. When you
add the fifth judge, the sixth judge, when you get as large as the old Fifth Circuit was, with
twenty-six judges, the law becomes extremely unstable. One of several thousand different
panel combinations will decide the case, will interpret the law. Even if the court has a rule, as
we did in the old Fifth, that one panel cannot overrule another, a court of twenty-six will still
produce irreconcilable statements of the law.'" Irving R. Kaufman, New Remedies for the
Next Century of Judicial Reform: Time as the Greatest Innovator, 57 Fordham L. Rev. 253,
259 (1988) (quoting Interview with Judge Gerald Bard Tjoflat, reprinted in 15 The Third
Branch, Apr., 1983, at 1, 3-4).
The other fundamental problem of an overly large court such as ours is with its en banc
process. The practical upper limit on the size of an efficient appellate court is that number of
judges who can effectively sit together en banc. Sitting together effectively requires three
things: (1) an oral argument in which the unstructured give and take between counsel and the
judges can accommodate the practical needs of all the judges to air difficult issues; (2) a
conference in which reasoned deliberation rather than mere voting can take place; (3) an
opinion writing process that can work the views of those judges in the majority into a majority
opinion rather than a plurality opinion. Other circuits sit together en banc with as many as 17
judges. The Fifth Circuit, before it split, decided that it could not, with fewer judges than the
28 we have on the Ninth. And the Ninth Circuit has decided by circuit rule to limit its en banc
court to 11 judges. Traditionally "en banc" used to mean the entire court. To the best of my
knowledge, we are the only appellate court in English common law tradition that calls less than
the entire court "en banc."
The word "collegiality" in its traditional meaning is critical to the en banc process. The
word is often used in contemporary speech to mean some combination of civility and
bonhomie. The traditional definition, though, is "shared authority among colleagues." The
American Heritage Dictionary, 291 (2nd College Edition, 1985). The word is derived from
"the doctrine that bishops collectively share collegiate authority." Id.
Judges betray their trust if they do not act with shared authority. We were not put in
office to be 28 individuals each imposing our idiosyncratic individual will on 50 million people.
The en banc process is what an appellate court uses to rein in those judges who may
mistakenly canter off the trail of established principles of law. But our en banc process does
not work, as the frequent Supreme Court reversals of our en banc dispositions show.
On a small court, the en banc process often works so well as a possibility that few en
bancs rehearings are needed. Opinions on some smaller courts are circulated to judges before
they are filed, they all have a chance to read them, and through internal memoranda within the
court, mistakes are corrected before mistaken opinions are even published. We cannot even
read each others' opinions after filing, so pre-circulation could not work on our large court.
On other courts, post-circulation en banc can assure that mistaken decisions are
repaired in accord with the views of the court as a whole. But not on ours. We have no true
en banc. Our en banc rehearings are before 11 judges. Eleven is just another panel, not even
40% of our 28 judge court. When an en banc panel is divided, the majority of the en banc
panel is not even a very substantial minority of the full court.
There is no reason to think that an en banc panel of eleven is representative of the
court. For one thing, no judge really represents another. We are not organized by party and
we are not representatives. We decide things individually.
Also, statistically, it is not especially likely that any particular en banc will be
representative of the court as a whole. An en banc panel is like a random draw of eleven cards
from a deck. It will rarely come out as a straight or flush. Even though a coin tossed enough
times will come out heads half the time, a single series of ten tosses (the chief judge is
automatically on the en banc panel) will not usually come out half heads. For example,
consider two judges on our court who are philosophically quite distinct, Judge Fletcher and
Judge O'Scannlain. From 1987 to 1997, ten full years in which they have both been on the
court, Judge Fletcher served on 62 en bancs, and Judge O'Scannlain on 34. Despite a random
draw, one served almost twice as often as the other on en banc panels! If you run a judicial
process as a crap shoot, than crap shooters' principles will affect the outcomes. The principle
here seems to be, "when you're hot, you're hot."
Benefits of a smaller circuit
A smaller circuit would not merely reduce the serious problems described above. It
would afford positive benefits. The judges could maintain much better familiarity with the law,
procedure, customs of the bar, and social and economic conditions in the different states within
Much federal law is not national in scope. Quite a lot of federal litigation arises out of
federal laws of only local applicability, such as the Bonneville Power Administration laws, the
laws regarding Hopi and Navaho relations, the Alaska National Interest Lands Conservation
Act, and the Alaska Native Claims Settlement Act. It is easy to make a mistake construing
these laws when unfamiliar with them, as we often are, or not interpreting them regularly, as
we never do.
Much federal procedure mirrors state procedure in the particular district. For example,
Federal Rule of Civil Procedure 4 imports state procedure. Where law is not specified, bar and
bench customs in the different localities often fill it in. It is very helpful for judges to know
how releases, attorney's fees contracts, and other documents for common transactions, are
typically written in a state, so that they know when something is suspicious and when it is
ordinary. In diversity cases, we are required to apply state law in federal court.
Yet on our court, ordinarily no judge on the panel has intimate familiarity with the law
and practices of the state in which the case arose, unless that state is California. A judge on
my court sits in Alaska perhaps once in ten years, and ordinarily never sits in Montana, Idaho,
Nevada, or Arizona.
Social conditions also vary, in ways that can color judges' reactions to facts, and disable
them from understanding the factual settings of cases not arising in California. For example,
judges from Los Angeles have different assumptions about what kind of people have guns than
judges from Idaho, Montana, and Alaska, who tend to associate gun ownership with a high
proportion, perhaps a considerable majority, of the longtime law-abiding residents of the state.
Native Americans have reservations in most states in our circuit, but in Alaska reservations
have generally been abolished. It is quite possible for Alaska lawyers not to point this out in a
brief because it is so obvious and well known, and for Ninth Circuit judges on a panel and their
law clerks, who have never been to Alaska, not to know it.
Finally, there is considerable benefit, for the development of good judicial
administration, to the diversity fostered by circuits that are not too large. We do not need a
rule that would work as well in Los Angeles as in Boise, if Boise and Los Angeles are in
different circuits, so it is possible for the procedural rules to be better fitted to the
circumstances to which they apply. A huge circuit such as ours is necessarily a procrustean
bed for much of the circuit.
Even where the problems are the same in different circuits, a diversity of approaches
can enlighten us as to what works best. For example, all the circuits are burdened with a very
large number of frivolous pro se appeals and petitions, and a very few meritorious and
important ones. All of us employ some form of screening to separate a few good ones from
the many bad ones. It is advantageous to the country to have more circuits trying different
approaches, so that we can discover from their varying experience what works best.
Another virtue of somewhat smaller circuits is that it limits the impact of our mistakes.
A certain amount of the time, we are going to be wrong.
Those who advocate more jumbo circuits like the Ninth Circuit overlook these costs.
Indeed, there is no logical limit to size, once these and the other costs of huge circuits are
disregarded. The logic of those who argue in favor of keeping the Ninth Circuit as it is
supports consolidating all of the circuit courts in the United States into a single giant circuit
court, which would function as an appellate pool from which panels of judges would be drawn
for the whole country. If you imagine the problems of maintaining coherence and predictability
of the law in such a circuit, you need only apply some discount to see what the problems are in
the Ninth Circuit.
But even if I am wrong on the causes of our errors, and even if our high error rate is a
temporary and chance occurrence, it is nevertheless undebatable that we do sometimes err.
And when we do, our error troubles one-fifth of the entire population of the United States. A
smaller circuit would limit the harm done by errors of that circuit.
Opponents of dividing the Ninth Circuit sometimes attack the motives of those who
favor a split, and some proponents argue in political terms as well. The arguments are
inappropriate. People can seek good things for bad reasons, and bad things for good reasons.
The regional distribution of judicial philosophies among members of the Ninth Circuit is
entirely fortuitous. Every President, with the advice and consent of the Senate, picks judges,
and he may find liberal Democratic lawyers in conservative Republican states, and conservative
Republican lawyers in liberal Democratic states. A Senator who criticized a recent decision we
made was proved right by a unanimous Supreme Court decision reversing us.
The second reason usually urged against a split is that it would cause "balkanization."
Balkanization is a metaphor. It suggests a condition like the Balkans. But in what sense are
the other circuits like Balkan states, Serbia, Croatia, and Bosnia? The First Circuit, the
smallest, has never been compared to Bosnia, and enjoys the greatest esteem. Balkanization is
just an over-excited metaphor for the normal process of different courts expressing different
points of view.
For most citizens served by our legal system, and the lawyers who serve them, it is not
very important whether the circuit they are in takes a different view of some issue of federal
law than another circuit, in which they do not live. They just need to know the law where they
live. What causes expense and complexity is differences of opinion between the federal and
state courts in the same jurisdiction, not divergence between the circuit where a litigant lives
and another where he does not.
Where inter-circuit conflicts do create problems, as in tax law, the problems can often
be repaired by issuance of regulations that clarify ambiguities, legislation, and, of course,
Supreme Court decisions. Supreme Court justices have frequently commented on the
usefulness to them of having issues "percolate" in the circuits, so that they can consider the
various view points that have developed along with their concrete applications, e.g., Arizona v.
Evans, 514 U.S. 1, 23-24 n.1 (Ginsburg, J., dissenting) ("We have in many instances
recognized that when frontier legal problems are presented, periods of 'percolation' in, and
diverse opinions from, state and federal appellate courts may yield a better informed and more
enduring final pronouncement by this Court." (citation omitted)).
For over 200 years we have seen the administrative benefits of dividing most of our
national governments by the number of states, and allocating the power to the separate states.
That too has given us the benefits of experimentation, as when a few states pioneered workers'
compensation laws. Judicial administration does not differ from other public administration in
this respect. What Justice Brandeis said of the benefits of dividing up public administration by
state also applies to our circuit courts. "It is one of the happy incidents of the federal system
that a single courageous State may, if its citizens choose, serve as a laboratory; and try novel
social and economic experiments without risk to the rest of the country." New State Ice Co. v.
Liebmann, 285 U.S. 262, 311 (1932) (Brandeis, J. dissenting).