Do school cliques dominate Japanese bureaucracies? Evidence from Supreme Court appointments.INTRODUCTION
Among American scholars, elite Japanese universities have a bad name. Forget how well the schools do or do not teach. Forget what research they do or do not produce. According to many American observers, they foster among their graduates a relentless exclusivity. Those graduates then form cliques, encourage their employers to hire even more graduates from their alma mater, and manipulate career tournaments to preserve favored posts for themselves.
To test this school-clique hypothesis, we need employee-level information on output: how much each employee produces. The elite university graduates did pass entrance examinations that others failed, after all. They might be smarter than their rivals. They might work harder. Before we can attribute any career success to cliques, we need to know the quantity and quality of the work that they do on the job. For most corporate and government positions, we have no such information.
Within the courts, arguably we do have that employee-level work product: we know the opinions a judge publishes. To test the school-clique hypothesis, I thus ask whether the judges from the elite universities enjoy more successful careers than their output would warrant. The quantity and quality of their opinions held constant, are they more likely to be named to the Supreme Court?
They are not. I find only weak evidence of any favoritism toward Kyoto University graduates, and no evidence of favoritism toward the graduates of the preeminent University of Tokyo. Elite university graduates do not dominate Supreme Court appointments because of their school backgrounds. They dominate because they produce.
In Part I, I summarize the American literature on Japanese school cliques. In Part II, I outline the structure of the Japanese courts; in Parts III.A and III.B, I summarize my data; and in Parts III.C and III.D, I report my results. In Part IV, I conclude by discussing some possible limitations, and in Part V, I discuss alternative measures of career success.
I. JAPANESE SCHOOL CLIQUES IN THE ACADEMIC IMAGINATION
A. The Possibility (1)
Whether in the American scholarly literature or in the Japanese newspapers, "school cliques" (known as "gakubatsu") dominated traditional Japan. They dominated firms. They dominated the government. And at least until some recent politically driven experiments, no clique dominated any place as thoroughly as the graduates of the University of Tokyo dominated the bureaucracy.
Elite Japanese universities select their students almost exclusively (the exceptions involve departments like physical education or the fine arts) through a blindly graded examination. Each school writes and administers its own. Some universities now cooperate on the first stage of an entrance examination. Even they, however, write their own distinctive--and determinative--second stage. Most universities write exams that test material mastered. A few (e.g., the University of Tokyo) write exams that test raw cognitive power.
Exam difficulty correlates with school prestige. The harder students find it to pass an exam, the higher everyone unofficially ranks the school. And the higher the rank, the more strongly employers compete to hire its graduates. Traditionally, the national University of Tokyo enjoyed preeminent status in nearly all academic departments. The national Kyoto University ranked second. A few national universities and private Tokyo-area schools filled the next tier.
According to American scholars (and commentators in the Japanese popular media), in the world beyond the university, the graduates of the elite schools look out for their own. They talk with each other. They mentor. They help. They lobby their employers to hire still more graduates. And they manipulate internal processes to promote fellow graduates over those from rival schools.
These school cliques, declares the late Berkeley and UC San Diego political scientist chalmers Johnson, constitute "without question the single most important influence within the Japanese state bureaucracy. The cliques of university classmates are inseparable from bureaucratic life ...." (2) Among the schools, none allegedly "does cliques" more effectively than the University of Tokyo. Explains Johnson, "[i]n place of the term gakubatsu, some Japanese analysts prefer Todaibatsu (cliques of Tokyo University classmates) because of the predominance of Tokyo University graduates in the bureaucracy and in the upper echelons of the banking and industrial worlds." (3)
To observers like Johnson, the cliques rig not just initial hiring decisions but later career moves, too. "Todai classmates in and out of government keep in touch with each other," he writes. (4) Tribal through and through, they are nothing if not corrupt. "once in the bureaucracy," declares Johnson, "the Todai group in an entering class in a ministry works together to ensure that its members prosper and that others are frozen out of choice positions." (5)
The late University of Washington legal scholar Dan Henderson echoes Johnson: the University of Tokyo graduates are successful, tribal, and successful because they are tribal. They "respect and promote each other's interests," he explains. (6) "[O]ne major irregularity evident in the high levels of the civil service is the favoritism (even clearer than in the hiring) shown for the Tokyo University (Todai) law graduates." (7) As evidence, he cites a study finding Tokyo graduates were "promoted faster (seven years on the average) and higher than law graduates from other universities." (8) As a consequence, "nearly 80 percent of the entire 'higher civil service' ... are Todai graduates." (9)
Sociologist B.C. Koh confirms the fact that University of Tokyo graduates thrive. Within government bureaucracies, he writes, "the proportion of Todai graduates is correlated with position level. That is to say, the higher the position level, the greater the proportion of Todai graduates." (10) Or consider, he explains, the Universities of Tokyo and Kyoto as a group. "The two universities together account for seven in ten higher civil servants overall, and their share of the pot increases to 89 percent at the bureau-chief level and to 95 percent at the vice-ministerial level." (11)
The courts constitute one such government bureaucracy, and many observers find similar cliques there. University of Tokyo cliques dominate the Ministry of Finance, and they allegedly dominate the courts. Among potential recruits, courts do seem to favor University of Tokyo graduates. In the private bar, fewer than 16% of all lawyers come from the University of Tokyo. (12) Of the 247 judges hired from 1959 to 1961, 23% were University of Tokyo graduates. (13)
And once in the courts, Tokyo and Kyoto graduates rise quickly to favored posts. (15) They spend more time in Tokyo and less in the provinces. They spend more time in prestigious assignments and less in branch offices. They control more powerful administrative posts and climb the pay scale more quickly. Among the twenty lower-court judges educated after the war and promoted to the Supreme Court by 2002, twelve graduated from the University of Tokyo and six from Kyoto. (16) Washington University legal scholar David Law similarly notes (and the data confirm) that the prime candidates for the supreme Court do tend to have attended the Universities of Tokyo or Kyoto. In the course of his discussion, Law focuses on the "grooming" that potential supreme Court appointees undergo:
At any given time, it will be possible to determine from [a given judge's] career to date whether he is a viable candidate for the Supreme Court. If he is in serious contention, he will have been groomed, or rewarded, with a series of assignments that place him firmly upon an elite career trajectory that would include many, if not most, of the following professional highlights. After compiling a distinguished academic career at the University of Tokyo (Todai) or Kyoto University (Kyodai), or possibly Chuo University, and achieving one of the top scores on the bar exam, he attends the LTRI and is then posted immediately or very soon thereafter to the Tokyo District Court. He will develop expertise in a particular area of law, be it civil, criminal, or administrative, and will at some point be tapped to serve as a law clerk, or chosakan, at the supreme Court. (17)
Law then elaborates at length on the type of other assignments elite judges routinely receive. (18)
B. The Puzzle
But do University of Tokyo graduates really rig the system? Many University of Tokyo graduates do enjoy spectacularly successful careers. Yet many also bring a spectacular reservoir of talent. Given that talent, they would receive attractive job offers whether the hiring was rigged or not. They would succeed in internal promotion tournaments whether rigged or not. And in truth, observers have never shown that Tokyo graduates actually rig procedures to favor each other anyway. They show simply that they outperform their competitors. Journalists then find passed-over employees from other schools who announce that their University of Tokyo rivals manipulated the tournaments that they lost, and American scholars repeat the claims.
The point is obvious, but perhaps worth stressing: University of Tokyo students passed the most selective university exam in the country. students do not pass it by accident. They pass it by combining extraordinarily high cognitive skills with a willingness to work relentlessly hard. They bring IQ and effort--and the two attributes are characteristics employers everywhere find valuable in the extreme.
As a result, the University of Tokyo graduates might simply do well because they are smart and work hard. They might do well on the job market because school cliques control hiring--but they might also do well because employers like smart and hard-working recruits. They might do well in the internal promotion tournaments because their clique controls the tournaments--but they might also do well because they outperform everyone else. (19)
Absent independent, employee-level data on work product, we cannot know. To tell whether cliques control hiring and promotions within Japanese organizations, we cannot rely on journalists. We cannot interview employees who wanted the posts University of Tokyo graduates took.
Instead, to tell whether university cliques control organizations, we need independent evidence on the quality and quantity of work that the graduates of the various schools perform. For virtually all firms and agencies, we will have no information on the output of individual workers. What is more, once an employer promotes one worker beyond his rivals, the members of his cohort will not be performing the same work anyway.
C. The Courts as a Test
In the courts, however, we may indeed have the data we need to compare output across employees. obviously, a government that rigs promotions in the Ministry of Finance will not necessarily rig them in the courts. Yet the empirical inquiry must start somewhere, and, in the courts, we arguably have the data we need to begin. A district judge is a district judge. He tries cases and decides them. Within any given district court (other than on specialized panels like intellectual property), he hears cases assigned to him randomly. The more able and hard working he is, the more cases he will handle and the less often he will be reversed.
From public records, I know the pace at which each Japanese judge climbs the career hierarchy. Generally, judges join the courts at the outset of their careers and stay until shortly before retirement age. From their job records, I can gauge their promotions.
Through other public records, I can also measure the quality and quantity of a judge's work. I know how many opinions the judge writes per year, and I know whether higher courts reverse them. The exercise raises obvious problems of selection bias (discussed in Part IV, below). But subject to several caveats, note that elsewhere, I similarly find that University of Tokyo graduates publish more opinions than graduates of other schools. (20)
II. THE JAPANESE COURT SYSTEM
Do judges who graduated from the University of Tokyo succeed in the career tournaments because their fellow graduates rig the administrative apparatus in their favor--as scholars like Johnson and Henderson imply? or do they succeed because they out-perform everyone else? To test the hypotheses, I ask which judges cap their careers with appointments to the Supreme Court. I first collect information on the backgrounds, productivity, and careers of a cohort of judges. I then contrast those who eventually became Supreme Court Justices with those who did not.
To check the robustness of the results, in Part V, I use the data for three other purposes: (a) to contrast judges who became High Court presidents with those who did not, (b) to contrast judges who became district court chief judges with those who did not, and (c) to contrast the University of Tokyo graduates with the graduates of other universities among the fast-track judges who began their careers at the Tokyo District Court.
A. Lower Courts (21)
Preliminarily, consider the architecture of the Japanese court system. Japanese judges work within a career bureaucracy. Where they toil, what they do, and how much they earn depend on how highly the officers in the court' s administrative office, the secretariat, regard their work (hence the claim that University of Tokyo graduates can rig the system). Those officers, in turn, are themselves career judges, albeit very successful ones. of the many posts at which a career judge can spend some time, service in the secretariat is one of the most prestigious.
The judges in the secretariat select the new judges that the court will hire. Nominally, the Cabinet appoints the judges, but in fact the Cabinet relies on the secretariat. The secretariat chooses its new recruits immediately after they graduate from the one national law school, the Legal Research and Training Institute (LRTI). Although critics urge it to hire practicing lawyers, to date, it has seldom done so.
The system by which students become lawyers, judges, and prosecutors recently changed in several ways. Because I compare judges who eventually became Supreme Court Justices with those who did not, I focus on judges hired several decades ago. As a result, the recent changes do not affect the discussion here.
The system goes as follows:
The LRTI ... admits students on the basis of a (mostly blindly graded) annual examination. During the period in question (the system recently changed), the pass rate on this exam hovered below 3 percent. Most people who took it never passed, and those who did typically passed only after failing it five or six times first. Students in Japan who would become lawyers, judges, or prosecutors usually studied law as an undergraduate subject. They then took the entrance examination to the LRTI. If they passed, they studied there for two years (recently changed). Upon graduation, they took jobs in private practice, on the bench, or in the prosecutorial office. Those who never passed typically worked in the legal departments of the large corporations. (22)
3. District and High Courts
"Most years, the Secretariat hire[s] 70 to 130 new judges a year. Over the course of their careers, these judges move through a series of appointments, generally at three-year intervals." (23) In the district courts, they hear cases alone, except for serious crimes, appeals from summary courts, and the more major civil cases. The latter group of cases they hear on three-judge panels. (24) Because court reporters disproportionately publish the more important cases, about two-thirds of the published opinions in district court civil cases involve three-judge panels. (25) The intermediate appellate courts are known as the high courts. These courts hear all cases as three-judge panels. (26) When judges hear cases on panels, the most senior judge structures the trial and determines the pace at which the panel will decide the case. (27)
"Virtually all [judges] spen[d] some time in courts considered undesirable, and virtually all also spen[d] time in coveted Tokyo or Osaka appointments. The more talented the judge, the more time he spen[ds] in urban courts" and prestigious administrative jobs like the Secretariat itself. (28) The more ordinary his abilities, the more years he spends in undesirable courts.
Lower court judges face mandatory retirement at sixty-five. shortly before turning sixty-five, a judge with respectable ability will typically find himself appointed chief judge to a district court. He will serve several years and then retire. A star will find himself appointed "president" (i.e., chief judge) of one of the seven high courts (i.e., intermediate appellate courts). A very select few will find themselves named to the Japanese Supreme Court. I discuss these appointments further in Part V.
B. Supreme Court (29)
Fifteen Justices serve on the Japanese Supreme Court. There, they hear cases either on five-judge panels or, exceptionally, en banc. They receive their appointments from the Prime Minister, usually when they are in their early sixties. They face retention elections from time to time, but no Justice has ever received a substantial negative vote. They serve until mandatory retirement at age seventy.
Of the fifteen Justices, by custom, the Prime Minister names five or six from the lower courts. The others bring backgrounds in the bureaucracy, the prosecutorial office, the bar, and the professoriate. Although lower court judges never write dissents, Supreme Court Justices may, but seldom do.
III. JUDICIAL PERFORMANCE
A. Introduction (30)
When the courts hire a new group of judges, the Secretariat can consult three types of information about each new hire's talent: (i) the selectivity of the university he attended; (ii) his year of birth, from which it can calculate how often he probably failed the LRTI exam; and (iii) the quality of his work product as reported by supervising judges, because LRTI students spend time as interns in the judiciary.
Traditionally, the secretariat took those new judges that it considered most talented and appointed them to the Tokyo District Court for their first three-year term (I consider this further in Part V.D.). Thereafter, it moved them through a variety of other courts and positions. As a result, an initial appointment to the Tokyo District Court signalled that the Secretariat had placed a judge on a "fast-track" within the courts.
I have some, but not all, of the information available to the Secretariat. For most (not all) judges, I know the university he attended (item (i)). I know a judge's age, and can estimate how often he failed the LRTI exam (item (ii)). And although I do not directly know the quality of his work product during his LRTI internship (item (iii)), I know where the secretariat started him. Given that the secretariat decides whether to start a judge at the Tokyo District Court on the basis of all three factors, I thus have an indirect measure even of a judge's performance at the LRTI.
B. Data and Variables
I take the information on a judge's tenure, background, and appointments from the fourth edition of the Zen saibankan keireki soran [Career Data on All Judges] (ZSKS), published by the Nihon minshu horitsuka kyokai in 2004. (31) The book is used routinely by observers of the Japanese courts. I know of no claims of systematic bias and no evidence of significant inaccuracies.
I obtain information on judicial output from the Hanrei taikei, the electronic database maintained by the Daiichi Hoki firm. Like Westlaw and Lexis, Hanrei taikei provides all published opinions in electronic form. (32) some of those opinions originally appeared in private commercially published reporters like the Hanrei jiho and Hanrei taimuzu. others appeared in reporters published by the courts.
Japanese district court judges write an opinion in all cases they decide. However, they do not decide all cases they handle, and the reporters do not publish all opinions they write. In 2000, for example, district court judges cleared about 1,194,000 civil cases. (33) Of those, about 187,000 were "litigation" cases. (34) Judges wrote decisions (hanketsu) in 80,542 of those civil litigation cases, (35) and the Hanrei taikei compiled 1,447 of the civil decisions. (36)
To obtain a cohort that reached retirement age by the publication of the fourth edition of the ZSKS in 2004, I examine all judges from the LRTI classes of 1959, 1960, and 1961. Note that a judge who turned twenty-four in 1960 would reach age sixty-five in 2001. Because this group produced only three Supreme Court Justices, I add career and productivity information on the four Justices appointed from the adjacent classes of 1957-58 and 1962-63.
Acquiring the information on the reversal rates and the time from filing to judgment for a judge' s opinions involves a more labor-intensive process. Accordingly, I collect this information only on judges from the LRTI class of 1960.
of the 247 judges in the classes of 1959-61 on which data is available, seven were women. None of the women were appointed to either the Supreme Court or the presidency of a high court. One served as chief judge to a district court. Although I include information on these women in this article, I do not focus on the implications of a judge's sex on his or her promotion. Instead, I discuss that issue in more detail in another Article. (37)
UNIVERSITY OF TOKYO: 1 if a judge graduated from the University of Tokyo, 0 otherwise.
KYOTO UNIVERSITY: 1 if a judge graduated from the Kyoto University, 0 otherwise.
OTHER UNIVERSITY: 1 if a judge did not graduate from either the University of Tokyo or Kyoto University, 0 if he did.
FLUNKS: the number of times a judge failed the entrance examination to the LRTI, estimated from his year of birth.
LOW FLUNKS: 1 if FLUNKS is 2 or fewer, 0 otherwise.
TDC START: 1 if a judge began his career at the Tokyo District Court, 0 otherwise.
SEX: 1 if a judge is male, 0 if female.
PRODUCTIVITY: the number of district court opinions published by a judge (both single-authored cases and those decided by three-judge panels), divided by the number of years he served on a district court.
TDC PRODUCTIVITY: the number of Tokyo District Court opinions published by a judge (both single-authored cases and those decided by three-judge panels), divided by the number of years he served on the Tokyo District Court.
PRIV RPTR PRODUCTIVITY: the number of district court opinions published by a judge (both single-authored cases and those decided by three-judge panels) in one of the two principal private law reporters (the Hanrei jiho or the Hanrei taimuzu), divided by the number of years he served on a district court.
TIME-TO-JUDGMENT: the number of years from the year a case is filed to the year of the district court decision (LRTI class of 1960 only).
REVERSAL RATE: the number of published opinions reversed by a higher court (in whole or in part), divided by the number of opinions published (LRTI class of 1960 only).
I include selected summary statistics in Table One.
C. Determinants of Productivity
Consider the proposition: (a) If universities, the LRTI, and the secretariat select students, lawyers, and judges for intelligence and effort, (b) if smarter and hard-working judges work more productively than others, and (c) if those smarter and harder working judges do not disproportionately promote out-of-court settlements, then University of Tokyo, Kyoto University, Flunks, and TDC Start should correlate with measured Productivity. They do. Table one shows that the correlation between Productivity and each of the four measures is .19, .03, -.18 and .26, with each significant at the 1% level other than Kyoto University.
Because of the low pass rate on the LRTI exam, most applicants never passed, and those who did pass did so only after many tries. one who passed while still in college would graduate from the LRTI at age twenty-four. Among the 247 judges hired from 1959 to 1961, only ten managed this feat. Fifty passed on their second try, and thirty-one passed on their third.
Students who pass selective university admissions tests also tend to pass the LRTI exam. Of the 247 judges hired from 1959 to 1961, fifty-six attended University of Tokyo and forty-three attended Kyoto University. The University of Tokyo graduates failed the LRTI exam 3.70 times, the Kyoto University graduates 3.91 times, and the other judges 4.36 times. The difference between the two elite schools and the others is significant at the 10% level. In the private sector, lawyers typically failed it 6.57 times. (38) of the ten judges who passed the exam on their first try, half had attended one of the two elite schools. of the sixty judges who passed it on one of their first two tries, 53% had attended one of the two.
Of the judges in these 1959-61 classes, twenty started their careers at the Tokyo District Court (the fast track). Among these Tokyo-starters, 45% had attended University of Tokyo and 15% had attended Kyoto University (the overrepresentation of University of Tokyo graduates is significant at the 1% level). The Tokyo District Court starters failed the LRTI exam 2.10 times; the rest failed it 4.31 times (the difference is significant at the 1% level).
In Table Two Regression (1), I regress (through probit) an initial appointment to the Tokyo District Court on a judge's university, and on the number of times he failed the LRTI exam. As the numbers above suggest, graduates of the University of Tokyo and judges who failed the LRTI exam the fewest times were most likely to start with one of these fast-track appointments to the Tokyo District Court.
The seven women in the classes of 1959-61 failed the LRTI exam a mean 4.43 times; none passed the LRTI exam on one of their first two tries. One had attended the University of Tokyo, and one had attended Kyoto University. None began her career at the Tokyo District Court.
2. Predicting Productivity
a. Basic Results
If the university and LRTI examinations measure cognitive abilities and levels of effort relevant to a judge's work (and if talented judges do not settle rather than decide cases), then--as noted immediately above-- the more talented judges (measured by these variables) should publish substantially more opinions. They do. (39) University of Tokyo graduates publish more than one-and-a-half times as many opinions as the others. More specifically, among the 1959-61 judges, Tokyo graduates published 2.30 opinions per year on the district court bench, while the rest published 1.54, as shown in Table Two (the difference is significant at the 1% percent level). Kyoto University graduates published 1.61 opinions per year (the difference is not significant).
Those who passed the LRTI exam on one of their first three tries published 2.13 opinions, while the others published 1.47 (the difference is significant at the 1% level). Those who started at the Tokyo District Court published 3.20 opinions, while the others published 1.58 (again significant at the 1% level). Parenthetically, note that men published 1.75 opinions per year, while women published 0.58 (significant at the 10% level).
In Table Two Regressions (2) and (3), I regress PRODUCTIVITY on these various background characteristics. TOKYO U, FLUNKS, and TDC START is each strongly significant.
b. Robustness Check 7 (41)
Perhaps, however, the higher publication rates for these elite judges reflect not their talent but their post. The logic proceeds in two steps. First, perhaps the Secretariat disproportionately appoints its favored judges (favored for whatever reason) to Tokyo. University of Tokyo graduates in the 1959-61 classes did spend a mean 4.84 years in the Tokyo District Court, for example, while the others spent only 2.81 years (the difference is significant at the 1% level).
Second, perhaps litigants disproportionately file the most newsworthy cases in the big cities. If so, then the case reporters will disproportionately publish cases from Tokyo. The 1959-61 judges did publish 2.88 opinions per year when in the Tokyo District Court, but only 1.71 opinions per year in district courts generally. (42) If the Secretariat appointed its most favored judges to Tokyo and the Tokyo courts heard the most interesting cases, then PRODUCTIVITY would correlate with the indices of favor even if the favored judges wrote no more opinions than anyone else.
This counterhypothesis does not hold. The judges with the conventional measures of talent published more opinions than the others, even within the Tokyo District court. Again, University of Tokyo graduates published one-and-a-half times as many opinions as the others, even if I limit the sample to judges serving on the Tokyo District Court. University of Tokyo judges published 3.75 opinions per year while on the Tokyo District Court; the others published 2.48 (43) (the difference is significant at the 10% level). The judges who passed the LRTI exam within three years published 3.34 opinions per year in the Tokyo District Court, while the others published 2.58 (not significant). And those who started at the Tokyo District Court published 4.10, while the others published 2.66 (not significant).
More rigorously, in Table Two Regression (4), I regress TDC Productivity on my measures of talent. Because only about half the judges spent time in the Tokyo District Court, the database is much smaller. In turn, this reduces the statistical significance of the results. Although the coefficients are no longer statistically significant, note that their signs are in the same direction. For the most part, the magnitudes of the coefficients are close to those in Regression (3) as well. Even among the judges in the Tokyo courts, the University of Tokyo graduates seem to publish more opinions than the rest.
c. Robustness Check II
Alternatively, perhaps the process by which trial opinions are selected for publication biases my numbers. Commercial legal reporters such as Hanrei jiho and Hanrei taimuzu publish some court opinions in Japan--namely, those that the editors think will sell subscriptions. official government reporters publish the rest. The way that the official reporters select their cases varies by court, but generally the judge who writes the opinion proposes it for publication to the local court's publication committee. Unless the committee thinks the opinion lacks precedential value, it approves it for publication. By some accounts, the local committees approve most publication requests.
Because of this process, Productivity will conflate quality and quantity. A judge with high measured Productivity did not just write many opinions. He wrote many opinions that the commercial editors and the local court publication committees thought were worth publishing. Obviously, this conflation of quality with quantity does not threaten the conclusions in this study. If anything, it instead strengthens my claim that the Japanese courts promote the highest-quality judges.
Hypothetically, however, the process by which the courts select opinions for their official government reporters could introduce a school-clique bias. Hypothetically, judges who graduated from the University of Tokyo on the local publication committee could try to favor their fellow Tokyo graduates by disproportionately selecting their opinions for publication. If so, then high-productivity figures would not reflect true productivity; they would simply reflect the school the judge attended.
To test this possible bias at the court publication committees, I construct PRIVATE REPORTER PRODUCTIVITY: the number of district court opinions a judge published in the two principal private commercial reporters, the Hanrei jiho and the Hanrei taimuzu, divided by the number of years he served on a district court. The editors of these reporters care only about selling magazines, not about favoring University of Tokyo or Kyoto University graduates. For the classes of 1959-61 judges, these two private reporters published almost exactly half of all published opinions.
In fact, the publication process does not bias my results. The correlation between productivity measured by those opinions published in the two private reporters and productivity measured by all other opinions is 0.57, significant at more than the 1% level. What is more, University of Tokyo graduates publish half again as much as the others, even within the two private law reporters. (44) The judges who passed the LRTI exam within three years published more than the rest, and so did those who started at the Tokyo District Court.
In Table Two Regression (5), I regress this PRIVATE REPORTER PRODUCTIVITY on the university variables, FLUNKS, and TDC START. FLUNKS loses statistical significance, but UNIVERSITY OF TOKYO and TDC START remain significant at more than the 1% level. Whether measured by all opinions or only by those in the private commercial reporters, Tokyo University graduates publish substantially more than the others. School bias at the court publication committees does not explain a judge's observed productivity.
D. Determinants of Supreme Court Appointment
The Prime Minister named judges to the Supreme Court who brought indices of talent already visible on the day it hired them. The judges came from prominent schools. Among the seven Justices from the classes of 1957-63, the Prime Minister appointed two from among the University of Tokyo alumni and four from the Kyoto University alumni. He appointed judges who had failed the LRTI exam a mean 1.00 times (the other judges failed it 4.17 times), and 71% of whom had begun their careers at the Tokyo District Court (only 7% of the other judges had).
2. Productivity: Summary Statistics
Although the judges named to the Supreme Court brought these obvious indices of talent, the Cabinet seems not to have relied on those indices. Instead, it appointed to the Supreme Court those judges who proved most productive on the bench. It did not favor University of Tokyo graduates because of their school backgrounds. Instead, it happened to name them only because it searched for the most productive judges, and Tokyo graduates were disproportionately among them.
I begin with some summary statistics. The Prime Minister named to the Supreme Court judges who had been spectacularly productive on the bench. The average judge not named to the Supreme Court published 1.66 opinions per year while on a district court. The typical University of Tokyo graduate published 2.30. The seven judges named to the Supreme Court averaged 6.36 opinions per year. Two of the seven published an unremarkable one to two opinions per year. The other five averaged between six and eleven. On the Tokyo District Court, these seven hyperproductive judges published 8.96 opinions per year. (45)
3. Productivity: Probit Regressions
In Table Three, I examine Supreme Court appointments more systematically. In each column, I regress a variable equal to 1 if a judge were appointed to the Supreme Court on a series of explanatory variables. For each regression, I give the marginal effect of the variable, followed by the absolute value of the z-statistic in parenthesis. In Regression (1), I regress the variables without a productivity measure. In Regression (2), I add District Court Productivity. Consistent with the robustness checks described above, in Regression (3), I add TDC Productivity, and in Regression (4), I add Private Reporter Productivity.
Crucially, the marginal effect of University of Tokyo is insignificant in all regressions. When I add productivity measures, Flunks becomes insignificant as well. Kyoto University remains weakly significant.
Instead of relying on these indices of talent observable at the outset of a judge' s career, the Prime Minister seems primarily to turn to measures of how effectively a judge actually worked. Other than a possible bias toward Kyoto University, he does not ask what school a judge attended. He does not ask how many times he flunked the LRTI exam. Instead, he asks how much work he accomplished on the bench.
Because none of the women on the lower courts was appointed to the Supreme Court, I cannot include Sex in the regressions. Note, however, that the least productive judge named to the court still published 1.15 opinions per year. The most productive woman published 1.81 opinions per year (but none during her nearly ten years on the Tokyo District Court). The other women published an average of less than one opinion per year.
4. Productivity: Rank Ordering
For a sense of the extent to which productivity matters, consider Table Four. Here, I reproduce selected data on the fifteen most productive judges in the dataset. Among the hyperproductive fifteen, Productivity ranges from 5.7 to 11 opinions per year--where the classes of 1959--61 averaged only 1.7. Symptomatic of the high performance of its graduates, seven of the fifteen (47%) had attended the University of Tokyo. Among the judges as a whole, only 23% had attended the university. All but two of fifteen had failed the LRTI exam three or fewer times, and all but four had failed it two or fewer times. Among the judges as a whole, the mean Flunks was 4.1.
Crucially, five of the seven Supreme Court Justices came from among the fifteen most productive judges. Although the sixth Justice, Shigeru Yamaguchi, averaged only 1.679 career opinions per year on the district court bench, during his 4.3 years on the Tokyo District Court, he averaged 6.923 opinions per year. By TDC PRODUCTIVITY, he ranked the eighteenth most productive judge. Obviously, even he could work fast when necessary.
A. Publication and Docket Clearance
I do not claim that the Prime Minister looks specifically at the number of decisions a judge publishes, and I have not heard any observers in Japan make that claim. Instead, he probably looks at variables correlated with that publication rate. Observers of the courts most commonly argue that the Secretariat promotes judges according to their docket-clearance rates. Probably, a judge's publication rate correlates with his ability to clear the docket.
Note that the cases that disputants choose to litigate are not a random sample of all the quarrels they fight, (48) and the opinions that reporters choose to publish are not a random sample of all opinions judges write. As noted earlier, in 2000, Japanese courts disposed of about 187,000 civil litigation cases. Judges wrote judgments in about 81,000 of those cases, and, according to the Hanrei taikei database, legal reporters published 1400 of those opinions. (49)
Hypothetically, judges who publish many opinions might not dispose of the largest number of cases. Nonetheless, note that a Tokyo University background, low Flunk scores, initial assignment to the Tokyo District Court, and appointment to the Supreme Court all correlate with high numbers of published opinions. In itself, this does not prove that publication rates also correlate positively with docket clearance rates. It does, however, provide indirect suggestive evidence for that proposition: Productivity predicts appointment to the Supreme Court because (as some observers claim) the Secretariat promotes judges on the basis in part of docket-clearance rates, and Productivity proxies for those rates.
B. The Effect of Delays
Curiously, although the courts promote judges with high measured Productivity, they do not favor judges who publish opinions with the shortest measured delays (Time-to-Judgment). Among the judges who joined the court in 1960, the future Supreme Court Justices decided their district court cases only slightly faster than the others: 2.15 years on average, rather than 2.43 years. The judges who passed the LRTI exam in fewer than three tries were slightly faster than the others (2.33 years rather than 2.50 years), while the University of Tokyo graduates were slightly slower (2.48 years rather than 2.40 years). Perhaps most important, none of these differences is statistically significant.
In fact, Time-to-Judgment and Productivity are correlated positively with a correlation coefficient of .20, significant at the 10% level. The more productive the judge, the longer the mean TIME-TO-JUDGMENT on his opinions. This is not as odd a result as it might initially appear: disproportionately, the low PRODUCTIVITY judges were "cream-skimming" judges who published the easy cases filed during their tenure, while the high PRODUCTIVITY judges were "house-cleaning" judges who published not just the cases filed during their time on the local bench but also a substantial number of older cases filed before they had even arrived. (50)
Consider the contrast between two roughly contemporaneous judges. Kunio Motoyoshi joined the court in 1960 and retired in 1996 for a position as a notary public. He compiled a record with both low PRODUCTIVITY, and low TIME-TO-JUDGMENT. He published seven opinions over the course of his fourteen years on the district court bench. Three of the opinions reported no filing date, but the other four he published expeditiously: a 1964 opinion in a case filed in 1964, a 1969 opinion in a case filed in 1967, a 1970 opinion in a case filed in 1969, and a 1972 opinion in a case filed in 1971.
By contrast, Akira Machida entered the courts in 1961 and joined the Supreme Court in 2001. He published massive numbers of opinions, and many of them were in cases that dated from the years before he joined the court. In 1962, his second year on the bench, he published nineteen opinions, one of them in a case dating from 1960, the year before he became a judge. In 1963, he published sixteen opinions, three of them from 1960. In 1964, he published twenty-seven opinions, two from 1960, two from 1959, and one from 1956. In 1965, he published another sixteen opinions, four from 1960, two from 1959, and one from 1957.
Unlike Motoyoshi, Machida did not just dispose of the cases filed under his watch. He cleared a substantial backlog on his court. Because that backlog included cases dating from the years before he became a judge, his opinions generate a high mean delay figure. He published cases with a long measured TIME-TO-JUDGMENT, in short, precisely because he accomplished so much work.
C. The Effect of Quality
Hypothetically, the most productive judges might sacrifice quality for quantity and make the most mistakes. In real life, they do not. Because Japanese courts do not (for the most part) cite other opinions, I can not measure quality by citation rates. At least by the cruder metric of Reversal Rates, however, the most productive judges did not cut quality: the correlation coefficient between Reversal Rates and Productivity is an insignificant .07.
The Reversal Rates of the 1960 judges bound for the Supreme Court do not differ significantly from those of the others (5.4% compared to 4.3%; not significant). The Reversal Rates for judges who failed the LRTI exam fewer than three times do not differ significantly from those of the others (5.4% compared to 3.3%; not significant), and neither do those of the judges who began their careers on the Tokyo District Court (4.2% compared to 4.3%; not significant). University of Tokyo graduates do enjoy slightly lower Reversal Rates (2.4% rather than 5.2%), but the difference is just barely significant at the 10% level.
V. OTHER APPOINTMENTS
Given how few judges end their careers on the Supreme Court, as a measure of school-clique influence, the test presents a problem of small numbers. Consider, therefore, two alternative measures of career success: high court presidencies, and district court chief judgeships. (51) Both are capstone appointments for successful judges but more common than an appointment to the Supreme Court. Among the 247 judges from the classes of 1959-61, three became Supreme Court justices. Eleven became high court presidents (but not Supreme Court Justices), and seventy-two became district court chief judges (but not Supreme Court Justices or high court presidents).
As still another measure of school-clique influence, consider initial entry onto the judicial fast track: a starting appointment to the Tokyo District Court. The Secretariat starts its most promising judges (8.1% of the 1959-61 cohort) at this court. In Part V.D. below, I ask whether the Secretariat favors University of Tokyo graduates when it appoints judges to this track.
B. High Court Presidents
The eleven judges in the 1959-61 cohort who became high court presidents (but not Supreme Court Justices) were talented men. (53) Of the eleven, seven had attended the University of Tokyo (only 20.6% of the other judges attended the school), and three attended Kyoto University (16.7% of the others). They had a mean Flunks of 1.273, compared to 4.307 for the others (significant at the 1% level). Seven of the eleven had started their careers at the Tokyo District Court, compared to 4.3% of the others (significant at the 1% level).
Although the eleven high court presidents published opinions, they were not spectacularly productive. Recall that the men who became Supreme Court Justices published 6.362 opinions per year on the District Court bench. The eleven who became high court presidents published 1.906 opinions per year. The rest of the bench published 1.647 opinions per year. The high court presidents published more than the other judges, but not statistically significantly so.
Nor is lackluster productivity of the high court presidents peculiar to the measure used. While on the Tokyo District Court, the high court presidents published 2.099 opinions per year. The other judges (I exclude the three who became Supreme Court Justices) published 2.779 opinions per year, which is more than the presidents. In the principal private law reporters, the future presidents published more than the others--1.206 opinions per year compared to 0.879 opinions per year--but the difference is not statistically significant.
Because the high court presidents brought very high indices of talent but only modestly high measured Productivity, regression analogous to that in Table Two yields significant coefficients on the talent variables but not on Productivity. (54) The result is obviously consistent with a story of school cliques. It is also, however, consistent with omitted variables: the possibility that the Secretariat may be promoting judges on the basis of a variable (like docket clearance rate) that correlates only imperfectly with my Productivity measure. If it happens not to correlate in the case of these eleven high court presidents, then the talent variables will acquire statistical significance in its stead.
C. District Court Chief Judges
Of the 247 judges in the 1959-61 classes, only three became Supreme Court Justices. Only eleven became high court presidents. A full seventy-two became district court chief judges. Precisely because over a quarter of the judges receive the appointment, it lacks the prestige of the other two capstone positions. For exactly that reason, however, it also offers a more statistically reliable test of the impact of any school cliques.
The judges who became district court chief judges started their careers with observable measures of talent. Of the seventy-two future chief judges, 30.6% graduated from the University of Tokyo. By contrast, 57.1% of the Supreme Court Justices and high court presidents had attended the school, but only 16.1% of those who finished their careers without any of these capstone appointments. (55) Of the seventy-two, 16.7% percent had graduated from Kyoto University, which is nearly identical to the fraction among the non-capstone judges. The seventy-two future chief judges had a mean Flunks score of 2.736, the Supreme Court Justices and high court presidents had a mean of Flunks of 1.214, and the non-capstone judges had a mean 5.019.
The chief judges were also productive. Where the Supreme Court Justices and high court presidents published 2.797 opinions per year while on a district court, the chief judges published 2.653 opinions per year. The other (non-capstone) judges had a measured Productivity of 1.186 opinions per year. At the Tokyo District Court, the future Justices and presidents published 3.821 opinions per year. The future chief judges published 4.342 opinions per year, but the non-capstone judges published only 1.629 opinions per year. With the two private reporters, the Justices and presidents published 1.633 opinions per year, while the chief judges published 1.550 opinions per year and the non-chief judges only 0.571 opinions per year.
Table Four presents much the same message. Of the fifteen most productive judges in the dataset, every one received a chief judge appointment before he retired. Conversely, among the forty least productive judges in the data set, only four became chief judges.
Given these numbers, one would not expect a regression to show a strong school-clique effect, and it does not. In Table Five Regression (3), I regress appointment to a district court chief judgeship on the university variables: Flunks, TDC Start, and Productivity. Productivity and
Flunks are both strongly significant. Kyoto University is insignificant, and the University of Tokyo is just barely significant at the 10% level. This last university effect hinges on the productivity measure used. If I use TDC Productivity, the marginal effect of the University of Tokyo becomes a statistically insignificant .114 (z-statistic of 0.94), while the productivity measure remains strongly significant at .068 (z-statistic of 3.35). If I use Private Reporter Productivity (arguably a stronger measure of quality than Productivity, since it reflects the decision of an independent journal to publish the opinion), the marginal effect of University of Tokyo falls to an insignificant 1.581 (z-statistic of 1.62), while the productivity measure remains strongly significant at .213 (z-statistic of 4.70).
D. Initial Tokyo District Court Appointments
Consider an alternative question: whom does the Secretariat name to the prestigious fast track starting appointments at the Tokyo District Court? Suppose school cliques rigged the appointment. If they did, then the University of Tokyo graduates who started their careers on this fast track would have been less talented than the other judges who started on the same track. Over the course of their careers, they would have published fewer opinions. Did they?
In fact, the opposite is true: over the course of their careers, the University of Tokyo graduates who started at the Tokyo District Court published more opinions per year (albeit not statistically significantly so) than the other judges who started at the same court. The nine University of Tokyo graduates published a mean 4.050 opinions per year. The eleven other judges published only 2.500 opinions per year. During their various stints on the Tokyo District Court during the course of their careers, the University of Tokyo graduates published 5.066 opinions per year, while the others published 3.314 opinions per year. With the private reporters, the Tokyo graduates published 2.427 opinions per year, while the others published 1.326 opinions per year.
In short, the Secretariat did not discriminate in favor of the University of Tokyo graduates when it selected new judges for the fast track. If anything, it seems to have worried about media accounts of bias and discriminated against the University of Tokyo graduates. It appointed them to the court only if they showed promise of becoming more productive than the others.
American scholars routinely attribute university cliques to Japanese firms and bureaucracies. Disaffected employees from other schools blame the cliques for their own career setbacks, newspapers repeat the claims, and American scholars take their interviews and the newspaper accounts at face value. The graduates of the most selective universities dominate their employers, they write. They cause it to hire more alumni from their alma mater. They manipulate the internal career tournaments to favor each other over the employees from rival schools.
For most employers, we lack the employee-level measures of output we need to test this school-clique hypothesis. For the courts, however, we have it. I take data on judicial output. I then ask whether judges from the most selective schools are more likely--holding output constant--to end their careers on the Supreme Court. For the most part, they are not. Although graduates from the most elite schools do capture a significant fraction of the Supreme Court seats, they do not capture those seats because of their school backgrounds. Primarily, they capture them because they accomplish so much work.
(1.) On traditional legal training in Japan, see J. Mark Ramseyer & Minoru Nakazato, Japanese Law: An Economic Approach 1-21 (1999) and Minoru Nakazato, J. Mark Ramseyer & Eric B. Rasmusen, The Industrial Organization of the Japanese Bar: Levels and Determinants of Attorney Income, 7 J. Empirical Legal Stud. 460 (2010).
(2.) Chalmers Johnson, MITI and the Japanese Miracle: The Growth of Industrial Policy, 1925-1975, at 57 (1982).
(3.) Id. at 59.
(4.) Id. at 60.
(5.) Id. at 62.
(6.) Dan Fenno Henderson, Foreign Enterprise in Japan: Laws & Policies 209 (1973).
(7.) Id. at 211.
(10.) B.C. Koh, Japan's Administrative Elite 139-40 (1989).
(11.) Id. at 142.
(12.) Nakazato, Ramseyer & Rasmusen, supra note 1, at 471 tbl.3.
(13.) See infra Table 1.
(14.) Sources: HANREI TAIKEI [COMPENDIUM OF CASES] (Daiichi Hoki CD-ROM, 2010); ZEN SAIBANKAN KEIREKI SORAN [CAREER DATA ON ALL JUDGES] (Nihon minshu horitsuka kyokai, 4th ed. 2004).
(15.) See J. Mark Ramseyer & Eric B. Rasmusen, Measuring Judicial Independence: The Political Economy of Judging in Japan 12-13 (2003) [hereinafter Ramseyer & Rasmusen, Judicial Independence]; J. Mark Ramseyer, Predicting Court Outcomes Through Political Preferences: The Japanese Supreme Court and the Chaos of 1993, 58 Duke L.J. 1557, 1568 (2009) [hereinafter Ramseyer, Predicting Court Outcomes]; J. Mark Ramseyer, Sex Bias in the Japanese Courts, in Empirical Studies of Judicial Systems 2008, at 197, 202 (Kuo-Chang Huang ed., 2009) [hereinafter Ramseyer, Sex Bias]; J. Mark Ramseyer, Talent Matters: Judicial Productivity and Speed in Japan, 31 Int'l Rev. L. & Econ. (forthcoming 2011) (draft dated Oct. 2010) (on file with the Washington University Law Review) [hereinafter Ramseyer, Talent Matters]; J. Mark Ramseyer & Eric B. Rasmusen, The Case for Managed Judges: Learning from Japan After the Political Upheaval of 1993, 154 U. Pa. L. Rev. 1879, 1890 (2006) [hereinafter Ramseyer & Rasmusen, Managed Judges].
(16.) Zen saibankan keireki soran [Career Data on All Judges] (Nihon minshu horitsuka kyokai, 4th ed. 2004).
(17.) David S. Law, The Anatomy of a Conservative Court: Judicial Review in Japan, 87 Tex. L. Rev. 1545, 1557 (2009) (footnotes omitted).
(19.) scholars in sociology and elsewhere have accumulated an impressive amount of scholarship consistent with the claim that employees tied to social networks are more productive than others. See, e.g., Mark S. Granovetter, The Strength of Weak Ties, 78 Am. J. Soc. 1360 (1973). Given both that judges work either alone or on three-judge panels, and that the assignment of cases to a judge within a given court is generally random, I do not see how ties to any social network would increase the productivity of a Japanese judge. This literature may well apply in some situations; I do not see why it would apply here.
(20.) Ramseyer, Talent Matters, supra note 15.
(21.) This general introductory material is taken from Ramseyer & Rasmusen, Judicial Independence, supra note 15, at 7-16; Ramseyer, Predicting Court Outcomes, supra note 15, at 1563-69; Ramseyer, Sex Bias, supra note 15, at 201; Ramseyer, Talent Matters, supra note 15; and Ramseyer & Rasmusen, Managed Judges, supra note 15, at 1881-89.
(22.) Ramseyer, Talent Matters, supra note 15, at 3-4.
(23.) Id. at 4.
(24.) Saibansho ho [Courts Act], Law No. 59 of 1947, art. 26 (Japan).
(25.) This figure is based on cases decided in early 2000.
(26.) Saibansho ho [Courts Act], Law No. 59 of 1947, art. 18 (Japan).
(27.) Ramseyer, Talent Matters, supra note 15, at 11.
(28.) Id. at 4.
(29.) For a fuller discussion of appointments to the Japanese Supreme Court, see Hiroshi Itoh, The Supreme Court and Benign Elite Democracy in Japan 197-240 (2010), and Ramseyer, Predicting Court Outcomes, supra note 15.
(30.) See generally Ramseyer & Rasmusen, Judicial Independence, supra note 15.
(31.) See Zen saibankan keireki soran, supra note 16.
(32.) Hanrei taikei [All Judicial Cases] (Daiichi Hoki CD-ROM, 2010).
(33.) Shiho tokei nempo, 1-Minji, gyosei hen [Annual Report of Judicial Statistics, 1Civil and Administrative], at tbl.1 (Saiko saiban sho jimu so kyoku, 2005).
(35.) Shiho tokei nempo, 1-Minji, gyosei hen [Annual Report of Judicial Statistics, 1Civil and Administrative], at tbl.20 (Saiko saiban sho jimu so kyoku, 2000).
(36.) See supra note 32.
(37.) Ramseyer, Sex Bias, supra note 15.
(38.) Nakazato, Ramseyer & Rasmusen, supra note 1, at 474.
(39.) Using a different data set and focusing on the most senior judge of a panel (the judge with the responsibility for trial management), Ramseyer finds that judges from elite university backgrounds and judges who passed the LRTI exam quickly publish substantially more opinions than the others. Ramseyer, Talent Matters, supra note 15.
(41.) I use the same test, and obtain the same result, in an earlier study based on a completely separate data set of medical malpractice opinions. See Ramseyer, Talent Matters, supra note 15.
(42.) See supra Table 1.
(43.) See supra Table 2.
(44.) See supra Table 3.
(45.) I focus on these seven because they are rough contemporaries of the three classes on which I have aggregate data. If (as seems likely) publication rates and practices changed over the years (the number of published opinions rose dramatically from 1950 to 1970), then comparing the measured Productivity of Supreme Court appointees spanning a longer period would not yield trustworthy results. Note, however, that by 2002, twenty judges educated after World War II had been appointed to the Supreme Court. Of the twenty, twelve had attended the University of Tokyo. The twenty had a mean Flunks of 1.95. The seven appointees used in the regressions had a measured Productivity of 7.81, while the other thirteen postwar appointees (most of whom had joined the courts before the seven others) had a measured Productivity of 2.89. The twenty appointees as a whole had a mean Productivity of 4.37.
(46.) See supra Table 1.
(47.) See supra Table 1.
(48.) George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Stud. 1 (1984).
(49.) See supra notes 33-34 and accompanying text.
(50.) The result also reflects simple measurement error. About two-thirds of all published district court opinions are the work of three-judge panels. As explained in Ramseyer, Talent Matters, supra note 15, the speed at which a panel decides a case reflects the efficiency of the most senior judge. Because I collect aggregate data on all opinions on which a judge was a panel member, my Time-to-Judgment figure reflects the efficiency of judges other than the one whose data I collect.
(51.) See infra Table 5.
(52.) See supra Table 1.
(53.) See supra Table 5.
(54.) See supra Table 5, Regression 2.
(55.) See supra Table 3.
J. MARK RAMSEYER, Mitsubishi Professor of Japanese Legal Studies, Harvard University. I gratefully acknowledge the helpful comments and suggestions of Louis Kaplow, David Law, Mark Levin, Tom Miles, Frances Rosenbluth, Hoger Spamann, and participants at workshops at the University of Chicago, Harvard University, and Washington University in St. Louis. I received generous financial assistance from the Harvard Law School.
TABLE ONE: SELECTED SUMMARY STATISTICS (CLASSES OF 1959-61) (14) A. Means and Medians: n Min Mean Median Max Tokyo U 247 0 .227 0 1 Kyoto U 247 0 .174 0 1 Flunks 245 0 4.131 3 31 Low Flunks (<3) 245 0 .371 0 1 TDC Start 245 0 .081 0 1 D Ct tenure 247 0 19.398 10.5 37.583 TDC Tenure 247 0 3.277 2.0 20.417 D Ct Productivity 243 0 1.714 1.254 11.027 TDC Productivity 131 0 2.882 1.400 20.000 Priv Rptr Prod'y 243 0 .922 .527 6.551 B. Correlation Coefficients (withp-values): TDC DC tcd Priv Rp Tokyo U Kyoto U Flunks Start Prod'y Prod'y Prod'y Tokyo U 1.00 Kyoto U -.25 1.00 (0.00) Flunks -.069 -.030 1.00 (0.28) (0.64) TDC Start .158 -.019 -.177 1.00 (0.01) (0.77) (0.01) DC Prod'y .185 -.028 -.176 .257 1.00 (0.00) (0.67) (0.01) (0.00) TDC Prod'y .161 -.012 -.069 .142 .843 1.00 (0.07) (0.89) (0.43) (0.11) (0.00) Pr Rp Pr'y .189 -.003 -.132 .252 .893 .778 1.00 (0.00) (0.96) (0.04) (0.00) (0.00) (0.00) TABLE TWO: PREDICTING FIRST APPOINTMENT AND PRODUCTIVITY (40) A. Regressions: (1) (2) (3) (4) (5) First Dist Court TDC Priv Rptr TDC Productivity Prod'y Prod'y Tokyo U .077 ** .868 *** .742 ** 1.210 .476 *** (2.08) (2.92) (2.50) (1.59) (2.59) Kyoto U -.0007 .173 .152 .309 .177 (0.02) (0.52) (0.46) (0.29) (0.87) Flunks -.020 *** -.087 ** -.070 ** -.013 -.031 (2.93) (2.50) (2.01) (0.10) (1.42) TDC Start 1.149 *** 1.064 .698 *** (2.74) (1.12) (2.68) n 245 209 209 122 209 Regression Probit OLS OLS OLS OLS Adj/Pseudo [R.sup.2] 0.13 0.06 0.09 0.00 0.08 Notes: Regression (1) gives the marginal effects of a probit regression. The parenthetical number below the coefficient gives the absolute value of the t or z statistics. ***, **, *: statistically significant at the 1, 5, and 10 percent levels, respectively. The judges are from the classes of 1959-61 only, and Regressions (2) through (5) include only those judges who stayed on the bench at least ten years. All regressions include a constant term. B. Selected Summary Statistics (Classes of 1959-61): Mean Mean Mean DC Prod'y TDC Prod'y Priv Rptr Prod'y Tokyo U grads 2.30 3.75 1.29 Kyoto U grads 1.61 2.77 0.92 Low Flunks 2.13 3.34 1.11 TDC Starters 3.20 4.10 1.82 Other University 1.41 2.42 0.78 (40.) See supra Table 1. TABLE THREE: PREDICTING SUPREME COURT APPOINTMENTS (46) A. Regressions: (1) (2) (3) (4) Appointment to Supreme Court Tokyo U .030 .007 .0003 .0009 (0.49) (0.41) (0.87) (0.27) Kyoto U .580 * .459 * 1.415 * .098 * (1.93) (1.77) (1.64) (1.73) Flunks -.029 * -.005 -.00001 -.002 (1.65) (1.26) (1.48) (1.53) TDC Start 1.781 *** .437 ** .017 * .180 ** (2.97) (2.19) (1.87) (2.28) Dist Ct Prod'y .003 ** (2.43) TDC Productivity .000007 * (1.88) Priv Rptr Prod'y .001 ** (2.28) n 2.16 213 126 213 Pseudo [R.sup.2] 0.48 0.60 0.66 0.58 Notes: Probit regressions giving marginal effects, multiplied by 100. Absolute value of the z statistics given below the coefficient. ***, **, *: statistically significant at the 1, 5, and 10 percent levels, respectively. Supreme Court Justices include Justices appointed from the classes of 1957-58, and 1962-63. All other judges are from the classes of 1959-61 only and include only those judges who stayed on the bench at least ten years. All regressions include a constant term. B. Selected Summary Statistics: Dist Ct Productivity Mean Tokyo n Flunks Univ Min Mean Max S Ct Justices 7 1.000 .286 1.156 6.362 10.887 High Ct Presidents 11 1.273 .636 0.900 1.906 4.552 (excl. S Ct Justices) Dist. Ct. Ch. Judges 72 2.736 .306 0 2.653 11.027 (excl. S Ct Justices or High Ct Pres's) All other judges 157-61 5.019 .161 0 1.186 4.934 TABLE FOUR: FIFTEEN MOST PRODUCTIVE JUDGES (47) Rank Name Class School Flunks Product'y DC CJ 1 Yashushi Tokioka 1959 1 11.027 Yes 2 Akira Machida 1961 U Tokyo 1 10.887 Yes 3 Takuji Izumi 1963 U Kyoto 0 10.345 Yes 4 Kazutoshi 1961 U Tokyo 4 8.276 Yes Yamamoto 5 Kaoru Yamashita 1959 U Tokyo 6 8.246 Yes 6 Masahiro Iseki 1961 U Kyoto 1 8.145 Yes 7 Toyozo Ueda 1963 U Tokyo 2 8.110 Yes 8 Sukeyasu 1959 U Tokyo 1 7.688 Yes Koizumi 9 Norio Yamamoto 1959 U Kyoto 1 6.857 Yes 10 Shoji Shinoda 1960 U Tokyo 2 6.471 Yes 11 Akira Watanabe 1959 3 6.464 Yes 12 Toshiaki Makino 1960 3 6.291 Yes 13 Masao Fujiii 1957 U Kyoto 1 6.203 Yes 14 Hiroharu 1959 Nagoya U 1 6.151 Yes Kitagawa 15 Tadashi 1960 U Tokyo 3 5.656 Yes Takahashi 95 Shigeru 1957 U Kyoto 1 1.679 Yes Yamaguchi 142 Toshihiro 1960 U Kyoto 2 1.156 Yes Kanatani Rank Name HCt SCt Pres 1 Yashushi Tokioka No No 2 Akira Machida Yes Yes 3 Takuji Izumi Yes Yes 4 Kazutoshi No No Yamamoto 5 Kaoru Yamashita No No 6 Masahiro Iseki No No 7 Toyozo Ueda Yes Yes 8 Sukeyasu No No Koizumi 9 Norio Yamamoto No No 10 Shoji Shinoda No No 11 Akira Watanabe No No 12 Toshiaki Makino No No 13 Masao Fujiii Yes Yes 14 Hiroharu Yes Yes Kitagawa 15 Tadashi No No Takahashi 95 Shigeru Yes Yes Yamaguchi 142 Toshihiro Yes Yes Kanatani TABLE FIVE: OTHER CAPSTONE APPOINTMENTS (52) (1) (2) (3) Appt to Appt to Appt to Sup Ct High Ct Pres Dist Ct Ch J Tokyo U .0001 .0070 ** .1666 * (0.41) (2.47) (1.69) Kyoto U .0046 * .005 * .0570 (1.77) (1.95) (0.54) Flunks .0001 .0002 ** .0845 *** (1.26) (2.21) (5.02) TDC Start .0044 ** .0346 *** -.0681 (2.19) (3.81) (0.40) Dist C Prod'y .00003 ** .0000 .1361 *** (2.43) (0.81) (4.28) n 213 206 195 Pseudo [R.sup.2] 0.60 0.57 0.27 Notes: Probit regressions giving marginal effects. Absolute value of the z statistics given below the coefficient. ***, **, *: statistically significant at the 1, 5, and 10 percent levels, respectively. Supreme Court Justices include Justices appointed from the classes of 1957-58, and 1962-63. All other judges are from the classes of 1959-61 only and include only those judges who stayed on the bench at least ten years. In Regression (2), I exclude the judges appointed to the Supreme Court. In Regression (3), I exclude those judges appointed either to the Supreme Court or to a high court presidency. All regressions include a constant term.