Second-Best Justice: The Virtues of Japanese Private Law 9780226282046

Citation preview

Second-Best Justice

Second-Best Justice The Virtues of Japanese Private Law

j. m a r k r a m s e y e r

the universit y of chicago press

chicago and london

j. mark ramseyer is the Mitsubishi Professor of Japanese Legal Studies at Harvard University Law School. The University of Chicago Press, Chicago 60637 The University of Chicago Press, Ltd., London © 2015 by The University of Chicago All rights reserved. Published 2015. Printed in the United States of America 24 23 22 21 20 19 18 17 16 15 1 2 3 4 5 isbn-13: 978- 0-226-28199-5 (cloth) isbn-13: 978- 0-226-28204- 6 (e-book) doi: 10.7208/chicago/9780226282046.001.0001 Library of Congress Cataloging-in-Publication Data Ramseyer, J. Mark, 1954– author. Second-best justice : the virtues of Japanese private law / J. Mark Ramseyer. pages cm Includes bibliographical references and index. isbn 978-0-226-28199-5 (cloth : alkaline paper) — isbn 978-0-226-28204-6 (ebook) 1. Civil law—Japan. 2. Liability (Law)—Japan. 3. Actions and defenses—Japan. I. Title. knx500.r36 2015 346.52—dc23 2015002788 This paper meets the requirements of ansi/niso z39.48–1992 (Permanence of Paper).

For Jenny, Geoff, and the BPPJH

As of 2013, the US population was about 307 million. The Japanese population was 128 million. $1.00 US equaled 118 Japanese yen in January 2015, 87 yen in January 2013, 93 yen in January 2010, 103 yen in January 2005, and 105 yen in January 2000.

Contents Acknowledgments xi chapter 1. Doing Well by Making Do

1

chapter 2. A Tort System That Works: Traffic Accidents 10 chapter 3. A System with Few Claims: Products Liability

35

chapter 4. Few Claims, but for a Different Reason: Medical Malpractice (I) 71 chapter 5. Medical Malpractice (II)

125

chapter 6. Wrong but Predictably Wrong: Labor, Landlord-Tenant, and Consumer Finance 164 chapter 7. A Second-Best Court chapter 8. Conclusion 239 Notes

243

References Index

281

261

206

Acknowledgments

I

ndirectly but crucially, I received the ideas in this book from my friends. I received some of the ideas from the books and articles they wrote. I received the rest from the wonderful conversations with them over the course of the last far-too-many decades. I thank them for all that they did to make this project possible. Everything worthwhile in this book comes from my friends; the rest is mine. More directly, many of these friends read and offered comments on the chapters in this book, and on the articles that formed the basis of those chapters. In alphabetical order, they include Jennifer Arlen, Wered Ben- Sade, Omri Ben- Shahar, Stephen Burbank, John Campbell, Yunchien Chang, Richard Epstein, Eric Feldman, Thomas Ginsburg, John Goldberg, Mark Grady, John Haley, Temple Jorden, Louis Kaplow, Ryotaro Kato, Daniel Kessler, Daniel Klerman, David Law, Robert Leflar, Mark Levin, Salil Mehra, Tom Miles, Curtis Milhaupt, Yoshiro Miwa, Sam Peltzman, Silviu Pitis, Dan Puchniak, Eric Rasmusen, Mathias Reimann, Marc Rodwin, Frances Rosenbluth, William Sage, Chiaki Sato, Steven Shavell, Holger Spamann, Kathryn Spier, Alan Stone, Frank Upham, Mark West, and Norma Wyse. I learned much from the chance to present the ideas in these chapters at stimulating workshops. In alphabetical order, the venues include the following: American Law and Economics Association, Bar-Ilan University, Conference on Empirical Legal Studies, Duke University, Emory University, Fordham University, Harvard University, University of Chicago, University of Michigan, University of Pennsylvania, University of Tokyo, and Washington University, St. Louis. I received very generous fi nancial assistance from the Faculty of Law at the University of Tokyo; and the East Asian Legal Studies Center, the

xii

Acknowledgments

John M. Olin Center for Law, Economics, and Business, and the PetrieFlom Center for Health Law Policy, Biotechnology, and Bioethics, at Harvard Law School. I received thoughtful and always considerate support from Deans Elena Kagan and Martha Minow. I wrote much of this book while under the gracious and kind hospitality of the University of Tokyo Law Faculty. On US medical malpractice law, I received excellent research assistance from Lindsey Beckett. I published earlier versions of some of this book in several fora. In each case, I thank the copyright holders for granting permission for their use here: “Bottom-Feeding at the Bar: Usury Law and ValueDissipating Attorneys in Japan,” in Wulf A. Kaal, ed., Festschrift zu Ehren von Christian Kirchner: Rect im Oekonomischen Kontext (Tuebingen: Mohr Siebeck Tuebingen, 2014), 135– 58; “Liability for Defective Products: Comparative Hypotheses and Evidence from Japan,” American Journal of Comparative Law 61 (2013): 617– 56; “Talent Matters: Judicial Productivity and Speed in Japan,” International Review of Law and Economics 32 (2012): 38–48; “The Effect of Universal Health Insurance on Malpractice Claims: The Japanese Experience,” Journal of Legal Analysis 2 (2010): 621– 86; “The Effect of Cost Suppression under Universal Health Insurance on the Allocation of Talent and the Development of Expertise: Cosmetic Surgery in Japan,” Journal of Law and Economics 52 (2009): 497– 522; “Universal Health Insurance and the Effect of Cost Containment on Mortality Rates: Strokes and Heart Attacks in Japan,” Journal of Empirical Legal Studies 6 (2009): 309–42; and “Do School Cliques Dominate Japanese Bureaucracies?: Evidence from Supreme Court Appointments,” University of Washington Law Review 88 (2011): 1681– 711. With Professor Eric B. Rasmusen, I coauthored “The Case for Managed Judges: Learning from Japan after the Political Upheaval of 1993,” University of Pennsylvania Law Review 154 (2006): 1879– 930, the basis for part of chapter 7. Professor Rasmusen very graciously gave permission to use some of the material from that article in this book. Figure 2.3 in chapter 2 is reprinted with the permission of Tokyo District Court Civil Traffic Accident Litigation Research Committee (Tokyo chisai minji kotsu sosho kenkyukai), and the K.K. Hanrei Times Sha.

Chapter One

Doing Well by Making Do

I

meet a lot of lawyers in this business. And for the most part, Japanese lawyers talk about the same things American lawyers do. Like American lawyers, they talk about their families. They talk about their colleagues’ families. They add details about their colleagues’ extramarital affairs. They complain about the compensation that their rival partners negotiate. They talk about the law too. They talk about new statutes. They talk about politically charged cases. They talk about opinions they think are wrong. They gossip about corrupt politicians. But some topics over which American lawyers obsess, Japanese lawyers never touch. They never wonder whether they can successfully game a state judge in a dubious product liability claim. They never try to convince juries to award damages their clients did not suffer. They never wonder whether to contribute to a judge’s reelection campaign. On many dimensions, the Japanese courts work far better than the courts in the United States, and the topics Japanese lawyers never touch over late-night drinks reflect some of the reasons why. At root, the Japanese courts do well because they try merely to “make do.” They do not aspire to perfection. They aspire only to “good enough.” And that is the subject of this book. *

*

*

2

Chapter One

Japanese fi le many fewer lawsuits than do Americans, and they fi le fewer suits for the same reasons no one would have this conversation in Japan. Those reasons lie at the heart of this book on “tort” disputes. I take as my implicit focus the contrast in litigation between the United States and Japan, and I take as my explicit focus the Japanese half of that contrast. On litigation in the United States, readers can turn to an enormous English-language library. I offer here a book on Japanese litigation. Implicitly, I write it to explore the US-Japan difference. Explicitly, I concentrate on Japan—and describe the way Japanese parties, lawyers, and judges handle some of the disputes that otherwise plague wealthy modern capitalist societies. People have long speculated about why Japanese fi le so few suits, but the content of their speculation has shifted over the years. In the 1960s and 1970s, they argued that Japanese fi led fewer suits because they lived in a culture that discouraged them from asserting their claims— because it simply was not something good Japanese did. The argument has largely disappeared. In part, it disappeared because Japanese do indeed assert their claims; they just do not assert them in court. And in part it disappeared because the argument usually gave circularity a bad name: explaining how people behave by citing their culture, while defi ning their culture by how they behave is not—as Talcott Parsons is said to have told class after class of sociology students—“terribly informative” (Geertz 1973). Since the 1980s, scholars have tended (only “tended,” the practice is not universal) to argue instead that Japanese fi le fewer suits because their legal system does not work. American scholars are not the only ones to make the claim. Japanese law professors make it routinely, too. Would-be plaintiffs in Japan face barriers across the board, they argue: the courts are too slow, crucial information is hard to obtain, and lawyers are expensive. They face other barriers specific to various legal fields: judges may impose too high a standard of proof or the statute may make impossible demands. The fi nger pointing goes in the wrong direction. Japanese do not fi le fewer suits because their legal system does not work. It works just fi ne. Courts are not slower than those in the United States. Information is not harder to obtain. Lawyers are not more expensive. Instead, we Americans fi le more suits because our system works so badly. Indeed, it works so badly that attorneys can make a living—a good living,

Doing Well by Making Do

3

at that—fi ling claims against fi rms with little evidence of wrongful conduct at all. File suit, and hire an expert to discover a wrong. Were the suit to go to trial, the plaintiffs would probably lose. But one cannot be sure, and the road to trial is long and paved with one toll charge after another. All too many fi rms fi nd it cheaper just to pay the lawyers to go away. Pay them they do, and the attorneys live to fi le again another day. *

*

*

Maybe we try too hard. Maybe if we tried to do a bit less, we could accomplish a bit more. Maybe our problems result from trying too hard to offer a perfect legal system. Japanese do not expect as much of their courts. They do not aspire to run a perfect system or to offer perfect justice. To be sure, interview a law professor, a judge, or a TV talk-show host and he will likely talk long and maybe even eloquently about his passionate commitment to truth and equity. But the government does not run a system that embodies any attempt to offer the best judicial system possible. It runs a system designed merely to make do. Economists call it the “theory of the second best” (Lipsey and Lancaster 1956). If something (the cost of the courts, e.g., or the inability of real-world judges to reconstruct the past) prevents us from obtaining exactly what we want on one dimension, sometimes we do better not to try to come as close as we can to that ideal. Sometimes we do better to scale back our hopes on other dimensions, too. Richard Epstein and Douglas Baird (Baird 2009, 13; Epstein 1995) recount a more intuitive version of the principle and attribute it to their late University of Chicago colleague Walter Blum. “Blum’s law,” they call it: “In law, 95 percent is perfection.” Sometimes, getting a problem “mostly right” gives better results than trying to get it “exactly right.” The difference between mostly and exactly right is not only the institutional cost—though those costs can be huge—but also the risk that the extra ambition will produce a result that is flatly worse. Our American legal system illustrates the phenomenon in spades in litigation over those private wrongs that I take as the focus of this book, wrongs we call “torts”—think barroom brawls, for instance, or traffic accidents, the exploding television set (a genre we call “products liability”),

4

Chapter One

and the operation on the wrong knee (medical malpractice). In many ways, in American courtrooms we adopt rules and procedures that imply we aspire toward 99 percent—and reap disasters. Had we made do with Blum’s 95, we might have earned better results all around. This contrast between “mostly right” and “exactly right” explains much of the contrast between US and Japanese litigation. To be sure, it does not explain all the contrast. It explains much, but in the chapters to follow I also detail other reasons for the contrast. Products liability litigation in the United States, for example, is also shaped by the fact that some state judges are elected, encouraging extortionate litigation with the ballot box in mind. Medical malpractice litigation in Japan is shaped by the way the Japanese national health insurance reduces the number of sophisticated state- of-the-art procedures that generate malpractice claims. Perhaps an example of the difference between “mostly right” and “exactly right” would help. With even routine disputes like fatal traffic accidents, we in the United States run trials in ways that make it hard for the people involved to predict what will happen. Fact fi nding, for instance, we assign to juries. Rather than to a professional who has seen it all before and whose past behavior lawyers can investigate, we assign the job to novices—and let them decide (almost) whatever they will. We adopt this largely unsupervised approach (“cull six people from DMV rolls, and take whatever you can convince them to give you”—my own words) in the name of individualized justice. Give all claimants the opportunity to tell their peers about the indignities they suffered and the injuries they sustained. Instead of assigning the job to a tired and cynical bureaucrat (called a “judge”), assign it to men and women without a professional bias (our “jury of our peers”). Rather than use evidence of what usually happens in similar situations, ask these ordinary men and women to consider each confl ict afresh. This devotion to individualized justice brings at least three pernicious effects. First, it makes it hard for a quarreling pair to stay out of court. Trials are expensive. All else equal, people would prefer to avoid them if they could. If they can agree about what a judge would do, they can indeed avoid them. Rather than litigate their fight, they can settle by reference to what a judge would say if they asked him. Unfortunately, the use of juries and the focus on the particularity of each dispute make outcomes less predictable—and when outcomes are

Doing Well by Making Do

5

unpredictable, people fi nd it harder to settle. To cut a mutually acceptable deal, they need to agree about what would likely happen if they litigated. Plaintiffs will settle if a defendant will pay at least as much as they would obtain (net of what they expect to pay their lawyer) in court. Defendants will pay if a plaintiff will take no more than what they would owe (including what they would pay their lawyer) in court. If the two parties hold similar estimates of the litigated outcome, they can settle and pocket the amount they would otherwise pay their attorneys. If they agree about what a judge would likely say if they asked him, they need not bother to ask at all. When outcomes are unpredictable, the odds increase that the two parties will hold (incompatibly) optimistic views of the litigated outcome. If that happens, the plaintiff will demand more than the defendant would ever pay. The defendant will refuse to pay even the least the plaintiff would ever accept. Lacking a “window for settlement,” the parties will have no choice but to litigate. Second, in some cases the American system introduces a distinct local bias. Juries in some counties famously favor local plaintiffs over outof-state defendants. In turn, lawyers fund a massive industry locating these counties. Because our procedural rules mostly let plaintiffs sue large corporations anywhere they like, lawyers fi le suit in the most biased county they can fi nd. Although most defendants in traffic accidents are local, the defendants in products liability claims are often out- of-state fi rms. In effect, if an enterprising lawyer can fi nd a plausible local victim in a plaintifffriendly county, he can sue the out- of-state defendant and reap a windfall. If he can aggregate multiple claims through procedural devices like the class action (yet another example of the American insistence on getting matters “exactly right” and insuring everyone “a day in court”), he can reap still more. Since the aggregated claims expose even big firms to the risk of bankruptcy, the fi rm may pay even attorneys with the weakest claims to desist. Third, in some classes of cases the cost of trying to get it “exactly right” can be particularly pernicious. Take medical malpractice. Although juries are clueless about medicine, we ask them to decide whether a doctor acted with appropriate care in arcane, high-tech procedures. Because we want to offer everyone a full recovery, we let plaintiffs prove a victim’s “pain and suffering.” And because we want to give everyone a

Chapter One

6

chance to prove his case, we let plaintiffs hire physician “experts” to invent stories about why they should recover amounts larger still. The collateral consequences can be huge. Those consequences go beyond the recovery in the individual case. When we let plaintiffs extort massive sums from manufacturers, sometimes the fi rms take the products off the market. When we let them extract unwarranted amounts from physicians, sometimes the doctors leave the state. *

*

*

Japanese courts do well by making do. They adopt little of this American individualized approach. Instead, they follow rules and procedures that suggest someone took seriously Blum’s rule and decided to accomplish more by attempting less. Japanese judges do not pretend to offer the level of particularized inquiry that we expect in American courts. In traffic accidents, they apply simple and public formulae to calculate a victim’s damages. They publish charts that explain the relative fault of the two parties in standard accidents. We do not live in Lake Woebegone. Most real-world victims are about average, so judges mostly award the damages and apply the rules that would be appropriate to the average case. Compensation is not perfect, and neither are the resulting incentives. But rather than get the numbers exactly right, they adopt strategies that are mostly right—and stop the inquiry. Because so much of what a Japanese judge will do is clear at the outset, the fraction of cases where the two parties are both optimistic about their chances falls. With less mutual optimism, the parties more often agree to an out- of- court settlement. In the process, they save the amount they would otherwise pay their lawyers. I illustrate these principles with tort disputes—the focus of this book. Traffic victims fi le many fewer suits than they do in the United States, but they do not “eat their losses” (chap. 2). Instead, they fi le demands against the wrongful driver. Those drivers then pay the victims out of court. They pay amounts that track the judgments the courts would impose if they refused to settle. They settle, in short, in the “shadow” of the litigated outcome. Japanese courts take the same concern for uniformity and predictability to products liability and medical malpractice disputes. They ap-

Doing Well by Making Do

7

ply no aggregation rules (class actions) for small claims. This obviously reduces the ability of a minor claimant to recover, but it also limits the ability of unprincipled lawyers to use fraudulent claims to extort settlements from risk-averse manufacturers. Japanese litigate fewer products liability claims than do Americans (chap. 3)—and the standardized tables from the traffic accident jurisprudence explain much of that story. Yet, although they also fi le fewer product liability claims out of court, they do not live with uncompensated wrongs. They fi le few claims because modern products are safe and defective products cause few injuries. Americans fi le more claims because American courts sometimes reward plaintiffs who fi le fraudulent claims and rarely punish their lawyers for facilitating the fraud. Japanese also fi le fewer medical malpractice claims (chaps. 4 and 5)—and again the standardized tables play a part. But Japanese also— again—fi le fewer claims out of court. They fi le fewer malpractice claims because they suffer fewer injuries caused by malpractice. They do not suffer less malpractice because they enjoy better medical care. They suffer less malpractice because they have worse medical care. The reason for the contrasting litigation patterns in medical malpractice, in other words, lies entirely outside the law of malpractice—and instead in the health insurance system. Japan maintains a national health insurance that dramatically reduces the level of care Japanese receive. To keep costs low, it reimburses complicated and sophisticated modern procedures at relatively low levels. Japanese doctors respond by focusing their practice on the simple routine procedures reimbursed at the relatively higher levels. Whether here or in Japan, malpractice claims stem from the most complicated, sophisticated, high-risk procedures. Patients do not fi le claims unless there is an adverse event, and (by defi nition) adverse outcomes result more frequently from high-risk procedures than low. What is more, ordinary human failings (e.g., a wandering mind) will more likely constitute “negligence” when the doctor is performing heart surgery than when he is prescribing a drug for asthma. The national insurance cuts the number of sophisticated, complex procedures—and, in the process, cuts the number of malpractice claims patients fi le. Japanese judges do make mistakes, sometimes serious (chap. 6). Given their populist sympathies and lack of economic training (a function, in turn, of university structures), they sometimes choose rules that let a claimant recover ex post, at the cost of slashing aggregate social

Chapter One

8

welfare ex ante. American judges make similar mistakes, of course. But the details of the mistakes differ. Japanese judges make it hard to discharge workers—slashing the incentive of employers to hire. They make it hard to evict tenants—slashing the incentive of landowners to provide rental housing. They let borrowers sue to recover “excessive” interest on consumer loans—slashing the incentive of fi nance companies to lend to the middle class. Even in their mistakes, however, Japanese judges make them predictably. Once courts adopt a rule, judges enforce it: ironically perhaps, precedent binds more tightly in civil-law Japan than in common-law United States. The same predictability that works so well in fields like traffic accidents can magnify the effect of mistakes judges make in others. Granted, the parties can sometimes mitigate the pernicious effect of a rule by taking it into account in their contract. Even when they cannot do that, they can at least use its predictable character to settle their dispute out- of- court. As Richard Epstein (1995, xii) put it, “permanence and stability are the cardinal virtues of the legal rules that make private innovation and public progress possible.” Japanese courts maintain this uniformity and predictability (with both its virtuous and dysfunctional effects) by ruthlessly controlling their judges (chap. 7). Japanese law graduates join the courts at the start of their career. In joining, they enter a tightly run bureaucracy. How well they do—the cities where they work, the posts they hold, and the pay they earn—depends on how highly their administrative supervisors (themselves also judges) evaluate their work. Those supervisors reward productivity and predictability. In response, Japanese judges tend to work hard and follow precedent. The uniformity and predictable character of the legal system follow. On the one hand, this produces a set of courts without flamboyantly brilliant jurists. The Japanese courts mostly lack the likes of Felix Frankfurter, Richard Posner, Benjamin Cardozo, or Learned Hand. On the other hand, the uniformity and predictability also minimize the eerily loony character of American torts. Manufacturers do not leave industries because of liability; obstetricians do not leave prefectures because of malpractice premiums; and automobile passengers do not obtain one million dollar– plus judgments from McDonalds over coffee they spill in their laps. *

*

*

Doing Well by Making Do

9

American courts set high goals and only haphazardly deliver: they aim for 99 percent, but sometimes accomplish very little. Japanese courts do well, by making do: they aim for 95 percent and mostly deliver that for which they aim. They hire capable professionals to handle adjudication: this is not the place for the jury of one’s peers or of state-by-state experimentation. They tell those professionals what to do: not only do they publish precedents but they also publish careful, detailed principles about how to resolve the common questions that judges encounter. And they monitor those professionals throughout their career: this is not the place for the independent judge who does what he thinks best. The result is a system that gives clear rules for generous recoveries. Stop by a bookstore and a college- educated plaintiff can buy a couple of books that will explain what he can expect from court. The rules are detailed, but clear and sensible. A college- educated defendant can do the same. Once the two have this information, they can settle out of court. Plaintiffs will collect standardized, uniform amounts, but these will be generous amounts and plaintiffs will collect them without paying much to a lawyer. Defendants will avoid the legal fees as well. Call it standardized, homogenized justice if you will. For the most part, it compensates victims uniformly, predictably, and cheaply.

Chapter Two

A Tort System That Works Traffic Accidents

T

raffic accidents matter. Traffic accidents matter both because they are so profoundly ordinary and because they can be so devastatingly tragic. They touch the lives of most people and destroy the lives of some. They are common; most are minor, but a few wreck cars, ruin careers, and kill. Traffic accidents matter to ordinary men and women because they can be some of the most catastrophic events in their lives. Traffic accidents matter in a way that captures a basic contrast in the way that the American and Japanese legal systems operate. When a distracted driver hits a cyclist in the United States, his wife will visit a law fi rm. Sometimes she will visit a courthouse. Rarely will she take their dispute through trial. She will almost never see the driver outside the courthouse. Before she and the driver settle, they will have retained lawyers, gathered evidence, and fi led papers in court. Before they settle, they will have burned enormous piles of money. In Japan, a dispute over an accident will usually take a different course—and the people affected will burn substantially less money. One Sunday fourteen years ago I left our house in Tokyo on my bicycle. I remember the early minutes of the ride, but nothing after those first minutes until I woke up in an ambulance. The next two weeks I spent in emergency care. At that point, I did consult a lawyer—but only out of excessive American zeal. Japanese cyclists can consult lawyers, too, if they want. Rarely do they want. The driver who hit me visited the hospital with a box of very expensive fruit (alas, I was on IVs). He arranged to pay the hospital fees (but my US insurance covered them anyway).

A Tort System That Works

11

With her husband in emergency care, my wife could have visited a bookstore (there are still a few left in Japan). There, she would have found the Japanese equivalent of “Traffic Accidents for Dummies.” In the book, she would have learned how to calculate my damages and how to estimate the driver’s relative negligence (Japanese tort law uses a “comparative negligence” regime). She would have learned approximately how much a judge would award us if we sued. It is all very simple—my lawyer gave us exactly that information in all of fi fteen minutes. With his own visit to a bookstore, the distracted driver might have learned the same information. Given that he and I had the same information about what would happen in court, we settled by reference to that amount on our own. Did the fruit and the hospital fees matter? Maybe—it was a nice box of fruit. But the driver and I had every reason to settle anyway. American cyclists and drivers are more likely to retain lawyers and visit courthouses because they are less likely to agree. Even with the famed American “discovery” mechanisms, they are less likely to reach similar estimates about what would happen in litigation. Unable to agree out of court about the probable litigated outcome, they have little choice but to pay their lawyers to ask the judge directly. The contrast comes in part from the way American judges and juries try to tailor the outcome to the peculiar circumstances of the specific driver and victim. Table how often they actually understand those individual idiosyncrasies. Japanese judges hardly even try. They vehemently doth protest to the contrary, of course. Japanese and American professors, judges, and lawyers all adopt the same sanctimonious rhetoric. But Japanese judges face professional incentives that cut against taking the rhetoric very seriously (see chap. 7). Where American judges aim for the particularized fi rst best, Japanese judges make do with the routinized second best. As a result, where American parties sometimes face stratospheric trial costs, Japanese parties encounter a routinized alternative that cheaply transfers funds from distracted drivers to their hospitalized victims.1 In the chapter that follows, I summarize the structure of the system. I begin with a discussion of traffic accidents generally (section I) and of their history in Japan (section II). I then turn to the resolution of traffic disputes: of both nonfatal (section III.B) and fatal (section III.C) accidents. I conclude by exploring the way judges determine comparative negligence (section IV).

12

Chapter Two

I. Torts and Traffic Accidents “Tort” and “contract” claims comprise a large part of what courts handle in any modern democracy. In Japan in 2011, of the cases fi led in district court, 275,000 were criminal and 744,000 civil. Of the 744,000 civil cases, 4,000 involved administrative or regulatory fights with the government. Of the rest of the civil cases, many concerned either torts or contracts (Saiko saibansho, Shiho tokei 2011). Classically, the distinction between contract and tort turned on whether the parties had cut a relevant deal in advance. In a contract claim, one of two parties argued that the other had refused to perform as promised. If a worker borrowed from a moneylender and failed to repay the funds, he breached his contract. In tort, one of the two argued that the other had harmed him in ways unrelated to any deal the pair may have made. If one punk beat another in a barroom brawl, he committed a tort. If a driver hit a pedestrian, he committed a tort. Traffic accidents constitute some of the most common torts in modern life. Unfortunately, although the Japanese courts counted traffic suits separately through the mid-1990s, they no longer publish the data. Take suits fi led over traffic accidents in either the district court or a summary court in 1994. Given the inherent lag in claiming, pair this 1994 litigation data with 1993 accident data. I have no reason to think the principles below have changed since the 1990s. Table 2.1 presents some of the results. In the mid-1990s, Japanese fi led 7,600 suits over 879,000 accidents. Among the forty-seven prefectures in Japan (the Japanese equivalent of the US states), the lowest fraction of suits per accident appears in Tochigi, a prefecture northeast of Tokyo with a population of two million. Tochigi had 53 suits and 18,165 accidents, for a ratio of .0029. The prefecture with the highest ratio was Tokyo, with its thirteen million population. Tokyo had 1,086 suits over 63,879 accidents, for a ratio of .017. In the United States, people fi led 7.4 million civil suits in state courts in 2005 (Langton and Cohen 2009, 1). On about two and a half times the population of Japan (314 million compared to 128 million), Americans fi led nearly ten times as many suits as the Japanese. Of the 7.4 million civil suits fi led, they settled 99.6 percent and took 27,000 to trial. Of the 27,000, 61 percent involved torts, and 35 percent (9,431) specifically in-

A Tort System That Works

13

Table 2.1 Litigation over traffi c accidents A. National level, 1994 suits and 1993 accidents Total traffic suits Accidents Deaths

7,577 878,633 10,972

B. Prefecture level, 1994 suits and 1993 accidents (n = 47)

Total suits Suits/injuries Suits/deaths

Min

25th

Mean

Median

16 .0029 .186

51 .0052 .359

161.2 .0079 .585

80 .0078 .461

75th 155 .0096 .629

Max 1,086 .0170 2.424

Sources: Keisatsu cho (1994); Saiko saibansho, Shiho tokei (1994). Notes: Suits are from 1994 and represent the sum of suits filed in the district courts and the summary courts. Given the inevitable lag in claiming, accident data from 1993 are used. Accidents are limited to those that involve at least some personal injury. The number of deaths differs slightly from those given elsewhere because of different inclusion rules by source.

volved a traffic accident (Langton and Cohen 2009, 2). In short, Americans actually tried more traffic cases than Japanese even fi led. 2

II. “Kamikaze Drivers” Happily, it is a phrase lost to history. Tokyo was not always the land of well-mannered and courteous (if occasionally gridlocked) cab drivers. In the late 1950s, it seemed instead the land of the suicidal professional. Visiting Americans called their cabbies “kamikaze drivers” and prayed hard as they careened through the streets. Among American cities, Tokyo’s streets most closely resemble those of Boston. They run a few blocks and then turn and stop. Most have no name, at best one lane, and corners that are usually blind. As in Boston, streets date from precar centuries when most people walked and a few led horses or cattle. The only Japanese cities with anything approaching a grid are Kyoto (modeled in the eighth century on the relentlessly symmetrical Tang capital) and Sapporo (settled in the nineteenth century). Tokyo residents walk relentlessly and ride bicycles regularly. The average Japanese takes 7,168 steps a day; the average American takes 5,117 (Bassett et al. 2010). Japanese own 72.5 million bicycles—.570 per capita; Americans own 100 million—.322 per capita. Those Dutch so routinely extolled as green own 16.5 million—.991 per capita (Top 10 Countries

14

Chapter Two

2013). Other than the weekend road racer in team-kit lycra, however, no one in Japan wears a helmet. Instead, bareheaded mothers load one infant on the front of their bicycle and one behind, and chat into their cell phones as they meander across the street. Helmets or no, public health numbers still favor the Japanese: traffic accidents kill substantially fewer Japanese (and Dutchmen) than Americans. In 2010, 32,885 Americans died in traffic accidents—10.7 per 100,000 population; 537 Dutch residents died—3.2 per 100,000; and 5,745 Japanese died—4.5 per 100,000. Accidents do not specifically kill fewer pedestrians in Japan. In 2010, 4,280 American pedestrians died in traffic accidents—1.4 per 100,000; 63 Dutch pedestrians died—0.4 per 100,000; and 1,987 Japanese died—1.6 per 100,000. And given that Japanese ride more bicycles than do Americans, accidents certainly do not kill fewer bicyclists in Japan. In 2010, 618 bicyclists died in the United States—0.20 per 100,000 population, .00618 per 1000 bicycles; 119 died in the Netherlands—0.72 per 100,000 population, .0071 per 1000 bicycles; and 929 died in Japan—0.73 per 100,000 population, .0128 per 1000 bicycles (Naikakufu 2012, 161; Top 10 Countries 2013). Things were worse in the late 1950s, when the “kamikaze” drivers screamed through these streets. The cabbies lacked any experience, and so did everyone else on the road. Not only did streets go nowhere, but also, in the late 1950s, sidewalks often did not exist. Instead, open sewers lined the narrow alleys. Many homes sported a sliding front door that opened immediately onto the tile covering the sewer. Suppose a sevenyear- old were late for school. He slammed open the front door, hopped over the sewer, and jumped straight into the street. Many of the kamikaze cabbies drove the sensibly boring midsized Crowns that Toyota designed for the domestic taxi market. In the mid1950s, few other Japanese owned any cars. Many did own a bicycle, and a few fastened crude one- cylinder engines to the rear wheel. As their incomes increased, they switched to motorcycles, most famously to the Honda “Super Cub.” At a time when the British Triumphs, BSAs, and Nortons sported 250– 750- cc engines, when Harley Davidson offered a choice between 900 cc and 1200 cc, and when even Briggs & Stratton put 100- cc engines on its push lawn mowers, the Honda Cub sported a tiny four-stroke 50- cc power plant. It was a glorious motorcycle for the masses. With a step-through frame, plastic fairings, and a gas tank tucked under the seat, women could drive the little machine in skirts. With its automatic clutch, a nov-

A Tort System That Works

15

ice could shift it whether he (or she) could handle a manual transmission. And with that 50- cc displacement, the novice could obtain an operator’s license without a driving test. By 1960, Honda sold over 500,000 50- cc Cubs, and Japanese owned some 670,000 under– 50- cc motorcycles (and motorized bicycles). Firms like Yamaha and Suzuki soon introduced look-alike competitors, and by 1965 Japanese owned nearly three million of the sub– 50- cc units (Kokudo kotsu sho 2013, 10). Japanese did not drop the Super Cub for Brando’s Triumph (think “The Wild One”) or Hunter Thompson’s BSA (think “Hell’s Angels”); instead, they kept the Cub and added the Subaru. Fuji Heavy Industries introduced the brand in the 1950s. Hitler may have planned Ferdinand Porsche’s Volkswagen “Bug” as the people’s car, but next to the Subaru the Bug looked positively limousinesque. The Bug at least displaced 1200- cc through Porsche’s four-stroke flat-four engine (much like the fi rst Porsche 356 engines). The 1958 Subaru used a 360- cc two-stroke twin (the East German Trabant would use a 600- cc two-stroke twin). The Bug was 4 meters long, and 1.5 meter high; the Subaru was 3 meters long and only 1.3 meter high (the Porsche 356 was 3.9 meters long, and the Trabant 3.4 meters). Rival Toyo kogyo (with its Mazda brand) bolted its own 360- cc engine to a three-wheeled frame and called it a truck. By the end of the 1960s, Fuji Heavy Industries had sold 390,000 Subaru 360s and begotten a wild mai kaa (“my car”) fad. As of 1961, Japanese owned 1.46 million of the sub– 360- cc cars and 364,000 larger cars. By 1966 they owned 3.1 million small cars and 1.7 million bigger ones. They continued to buy the smaller cars but increasingly opted for larger units, too: by 1971, they owned 6 million small cars and 6.6 million larger ones (Kokudo kotsu sho 2013, 2– 3). The mai kaa fever did not make for defensive driving, at least not for effective defensive driving. By 1970, traffic accidents killed nearly 17,000 Japanese (see table 2.2). On a population of 105 million, the toll came to 16.0 per 100,000 population. Given the 19 million cars and trucks in Japan, the 17,000 Japanese deaths came to .886 per 1,000 vehicles. Over the four decades since, matters improved unambiguously. By 2010, Japanese owned 79 million cars. For all the cars, they killed only 4,800—3.8 per 100,000 population and .062 per 1,000 vehicles. From 1970 to 2010, the number of vehicles had climbed fourfold. The number of accidents held steady, and the number of deaths fell by 70 percent. Two parallel shifts cut the number of traffic deaths: the number of accidents per car fell and the number of deaths per accident fell. From 1970

6.86 .038 3.18 .160 .886 .074 23.3

Accidents/population Accidents/automobiles Accidents/km Deaths/population Deaths/automobiles Deaths/km Deaths/accidents

4.23 .016 1.65 .096 .370 .038 22.8

111,940 29,143 286,345 472,938 10,792

1975

4.07 .012 1.23 .075 .225 .023 18.4

117,060 38,992 389,052 476,677 8,760

1980

4.57 .011 1.29 .077 .192 .022 16.8

121,049 48,240 428,442 552,788 9,261

1985

5.20 .011 1.02 .091 .186 .018 17.5

123,611 60,499 628,581 643,097 11,227

1990

6.07 .011 1.06 .085 .152 .015 14.0

125,570 70,107 720,283 761,789 10,679

1995

7.34 .012 1.20 .071 .120 .012 9.7

126,926 75,525 775,723 931,934 9,066

2000

7.31 .012 1.21 .054 .087 .009 7.4

127,768 78,992 768,879 933,828 6,871

2005

5.67 .002 0.97 .038 .062 .007 6.7

128,057 78,661 746,008 725,773 4,863

2010

Sources: Songai hoken (2012); Somu sho (2013a); Keisatsu cho, ed., Heisei 24 nenju no kotsujiko no tokucho oyobi doro kotsuho ihan torishimari jokyo ni tsuite [Regarding the Characteristics of Traffic Accidents and the Handling of Traffic Law Violations in 2012], 30 (February 14, 2013), http://www.e-stat.go.jp/SG1/estat/List.do?lid=000001106841. Notes: Population is per thousand; automobiles are per thousand and include trucks but not motorcycles; Km traveled is per one million; deaths are from National Policy Agency records and give the numbers of victims who die within thirty days of the accident.

104,665 18,919 226,017 718,080 16,765

Population Automobiles Km traveled Accidents Deaths

1970

Table 2.2 Traffi c safety over time

A Tort System That Works

17

to 2010, the number of accidents per 1,000 automobiles fell from .038 to .002. Some of the factors are easy to surmise: more traffic signals, sidewalks, overpasses, and experienced drivers. Among those accidents, the fraction (per 1,000 accidents) causing deaths fell from 23.3 to 6.7. Again, at least some of the reasons are easy to guess: seat belts, safety glass, airbags, energy absorbing front ends, disk brakes, and improved tires. Compared to other wealthy democracies, in some ways Japan now seems safer than most. To be sure, it reports more accidents: the Japanese rate of .0098 accidents per 1,000 automobiles is the highest among the countries represented in table 2.3, and so is the rate of .097 accidents for every 100 million kilometers traveled. Yet the concept of a reportable accident may vary from country to country. Perhaps Japan just reports a larger fraction of its accidents. The concept of death is less flexible, and Japan reports relatively low death rates. To be sure, .770 deaths per 100 million kilometers traveled is the highest among the countries on table 2.3. But .045 deaths per 100,000 population is lower than the rates in the United States, Canada, France, or Italy. The .078 deaths per 1,000 automobiles ratio is lower than the number in the United States, Canada, Germany, France, or Italy.

III. Recovering Damages A. Introduction Many accidents or few, once a driver hits someone, the people involved potentially turn to the legal system. When Grace Kelly spun her car off the Monaco cliffs, no one filed a wrongful death claim. If a pedestrian checking messages on his phone steps in front of a driver who could not have missed him, some pedestrians would file a claim—but certainly not all. But if the driver could have avoided the pedestrian if only he had paid attention himself, the injured pedestrian will likely demand compensation. Japanese overwhelmingly handle these traffic disputes out of court. On 761,000 accidents (Saiko saibansho, Shiho tokei 1995; table 2.2), they fi led only 7,600 suits. In the rest of this chapter, I explore what happens to their claims—not only the 7,600 suits fi led but also the more than 700,000 other claims. Do victims assert their legal interests? If they do, what do they collect? How do their recoveries vary between in- court and out- of- court resolution? I begin with nonfatal injuries (section B) and spend the rest of the chapter on deadly accidents.

.195 .0062 .0319 .107 .132 .678 21.3

Km/automobiles Accidents/automobiles Accidents/km Deaths/population Deaths/automobiles Deaths/km Deaths/accidents

.161 .0059 .0367 .065 .106 .659 17.9

34,109 20,771 3,351 123 2,207

Canada

.149 .0063 .0424 .045 .080 .537 12.7

81,802 45,499 6,792 288 3,648

Germany

.139 .0048 .0347 .031 .057 .414 11.9

61,789 33,161 4,607 160 1,905

UK

.139 .0018 .0123 .064 .106 .762 59.6

62,799 37,590 5,239 67 3,992

France

.170 .0033 .0195 .028 .055 .325 16.6

9,341 4,829 819 16 266

Sweden

19.4

.068 .108

.0056

211 4,090

60,340 37,911

Italy

.146 .0013 .0087 .032 .062 .426 48.82

16,575 8,651 1,262 11 537

Netherlands

.101 .0098 .097 .045 .078 .770 7.91

128,059 73,964 7,460 726 5,745

Japan

Source: Naikakufu (2012), 161; data for 2007–10, by item and country. Notes: Population is per thousand in line 1, and per thousand in line 9; automobiles are per thousand; km traveled is per hundred million; and accidents are per thousand. Values are for 2007–10, varying by item and by country.

308,745 248,609 48,531 1,546 32,885

Population Automobiles Km traveled Accidents Deaths

US

Table 2.3 Traffi c safety across borders

A Tort System That Works

19

B. When No One Dies Nonfatal accidents are a diverse bunch. In many, the drivers simply damage their cars. In others, they injure each other, passengers, or pedestrians. Unfortunately, the wide variety among these accidents makes it hard to generalize from the numbers. In some, a passenger breaks multiple bones, spends weeks in intensive care, and returns to work only after months of rehabilitation. In others, a driver complains of headaches for a couple of weeks, fi xes his car, and forgets the incident. With caveats about the diversity involved, consider some simple numbers. In 2010, Japanese drivers caused 725,773 accidents. They injured or killed 895,326 people (Naikaku fu 2012, 161). Liability insurers paid at least some amount in 5,452 cases where a victim died, in 64,022 cases where someone became disabled, and in 1,147,392 other cases involving personal injury (ibid., 80). The numbers present a simple moral: victims fi le claims and recover compensation. The numbers do not tell us whether the victims recovered much or little. They do tell us that victims assert their claims. And they tell us that they recover some amount. The puzzle is not why so few Japanese assert their legal interests—according to these numbers, virtually all victims assert claims. Instead, the puzzle is why insurers pay so many apparent victims: in a year when drivers injured 895,000 people, their insurers compensated 1.15 million people for injuries. The answer probably turns in part on the lag between injuries and compensation. As Japanese roads grew safer, the number of accident victims declined. From a high of 1.2 million in 2004, the number of accident victims fell monotonically to 859,000 in 2011 (Naikakufu 2012, 6). Given the inevitable delay in processing claims, insurers will always pay some claims from prior years. In a world of falling accident numbers, this will generate deceptively high claim payment rates.

C. When a Person Dies 1. frequency of recovery As with accidents more generally, so, too, those involving a death: plausible claimants assert their claims. And as with claimants more generally, so, too, the heirs to traffic victims: plausible claimants recover

Chapter Two

20

against liability insurers. Turn fi rst to table 2.4: the number of traffic deaths within thirty days of an accident (as reported by the National Police Agency), and the number of deaths for which a liability insurer paid some amount of compensation. Note that the latter ranges from 84 to 91 percent of the former. In 10–15 percent of fatal accidents, liability insurers pay nothing. In all the rest, they pay. This 84– 91 percent range is bizarrely high. Some people drive too fast, spin their cars off the highway, and die. In these single- car accidents, a life insurer may pay the victim’s heirs, but not a liability insurer. Others may drink too much and hit another car, but kill only themselves. Again, no liability insurer will pay the decedent’s heirs. Yet take the single- car accidents (table 2.5). In 2012, drivers caused 4,280 fatal accidents. Of these 4,280, 878 accidents involved only one car and another 45 involved a train. In neither category will a liability insurer pay compensation. These accidents, however, constituted 21.6 percent of the total. As with accidents more generally, the puzzle is not why so few heirs fi le claims—apparently, heirs always fi le them. Instead, the puzzle is why Table 2.4 Numbers of automobiles, compensated deaths, and total deaths

Year

Automobiles

Compensated deaths

Total deaths

Percentage compensated

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

36,903,078 37,101,038 37,535,545 38,159,188 38,106,586 37,648,994 38,492,877 38,590,102 38,533,759 38,373,670 38,731,246 38,378,882 39,067,723 38,674,832 38,791,770 41,775,207 38,565,312 38,674,100 38,206,667

11,063 10,703 10,773 10,492 10,197 9,595 9,413 8,935 8,456 8,341 7,866 7,277 6,807 6,168 6,029 5,482 5,128 4,922 4,777

13,269 12,768 12,670 11,674 11,254 10,805 10,372 10,403 10,060 9,575 8,878 8,492 7,931 7,272 6,639 6,023 5,772 5,745 5,450

83.37 83.83 85.03 89.87 90.61 88.80 90.75 85.89 84.06 87.11 88.60 85.69 85.83 84.82 90.81 91.02 88.84 85.67 87.65

Source: Songai hoken (2012). Notes: Compensated deaths give the number of deaths for which some liability insurance was paid (but exclude payments from the small number of mutual aid associations); total deaths are numbers of victims who die within thirty days of the accident, as reported by the National Police Agency.

A Tort System That Works

21

Table 2.5 Fatal accidents, by type of accident

A. Pedestrian B. Multivehicle C. Single vehicle D. Locomotive E. Total F. (C+D)/E

2006

2007

2008

2009

2010

2011

2012

1,482 2,815 1,333 45

1,401 2,512 1,181 35

1,244 2,312 1,005 44

1,204 2,101 1,020 37

1,238 2,046 1,018 35

1,171 1,949 908 38

1,141 1,790 878 45

6,196

5,625

5,067

4,826

4,783

4,532

4,280

22.2

21.6

20.7

21.9

22.0

20.9

21.6

Source: Keisatsu cho, ed., Heisei 24 nenju no kotsujiko no tokucho oyobi doro kotsuho ihan torishimari jokyo ni tsuite [Regarding the Characteristics of Traffic Accidents and the Handling of Traffic Law Violations in 2012], 30 (February 14, 2013), http://www.e-stat.go.jp/SG1/estat/List.do?lid=000001106841. Notes: Data are number of accidents involving a fatality (not the number of fatalities) for accidents (a) between a vehicle and a pedestrian, (b) between two or more vehicles, (c) involving only one vehicle, and (d) involving a train.

liability insurers nearly always pay. According to table 2.4, liability insurers pay the heirs to 84– 91 percent of all traffic accident victims. And according to table 2.5, 20– 23 percent of the fatal accidents generate no plausible liability insurance claim. Probably, the reason for the bizarrely high payout rate again turns on the lag between accident and payment. Given the steadily falling number of traffic deaths, insurers will always be paying some claims from prior years when more people died. The more basic moral, however, is again simple: heirs to victims of traffic accidents always fi le claims against the plausibly offending driver’s insurance company. 2. recovery formula Before turning to the amounts that heirs collect out of court, consider the formulae that judges follow in deciding how much to award. These formulae are not mandatory. Judges could ignore them if they wanted. They apply them to the cases they face because they work within a regimented bureaucracy in which their supervisors (other judges) reward them for doing exactly that: where they work, what they do, and how much they earn turn on how fast and how predictably they clear their docket (see chap. 7). In the typical wrongful death case, judges award amounts that include five components. 3 First, they award pain and suffering. Given that they mandate the amounts to compensate at least in part for the trauma suffered by the heirs rather than the decedent, the amounts

22

Chapter Two

vary by the role the decedent played in his family. Thus, by recent accounts (Ariyoshi et al. 2012, 80, 108; Usugane 2012, 99), if the heirs depended on the decedent’s earnings for income, the courts tend to award amounts in the 27– 31 million yen range. If the decedent played a central role in raising young children, or cared for aging parents or young siblings, they award 24– 27 million yen. For all other victims, they award 20– 25 million yen. Second, judges award present-valued lost wages, net of living expenses (Ariyoshi et al. 2012, 82–108; Usugane 2012:,78– 91). If the victim worked, they use his actual salary. If the victim was of working age but a mother who stayed home to care for her family, they use the mean salary for women her age. If the victim was still a child, they use the mean salary for employees of the child’s sex. And if the victim had already retired, they pay no lost wages. From the present-valued lost wages, judges deduct expected living expenses. They do not deduct actual amounts. Instead, by tradition they deducted 50 percent for single men. More recently, some courts have begun to deduct only 30–40 percent for those who left a dependent family. They are particularly apt to deduct the lower fraction when women die, lest they value female lives at lower prices given the generally lower wages women earn in Japan. Third, judges award funeral expenses. Rather than award actual expenses, they tend to award a straight 1.3–1.7 million yen (Ariyoshi et al. 2012, 81; Usugane 2012, 46–47). Fourth, judges award out- of-pocket expenses (Ariyoshi et al. 2012, 81). Where the victim or his heirs incurred medical (or other) expenses between the time of the accident and his death, they award actual expenses. Last, judges award attorneys fees. They award them as part of a oneway fee-shifting regime: plaintiffs recover fees if they win but not if they lose, and defendants do not recover fees even if they win. Rather than award actual amounts, according to popular wisdom, judges tend simply to add about 10 percent to the judgment.4 3. actual litigated recoveries (a) Summary To investigate the value that judges actually attach to human life, I take all opinions published from 2005 to 2009 that concerned a fatal traffic

A Tort System That Works

23

accident. I locate the opinions with the Hanrei taikei database. A Japanese Lexis-Westlaw equivalent, the database purports to include all reported opinions. Through this search, I identify 130 opinions over the five-year period, from 21 opinions in 2006 to 35 in 2009. In these published opinions, judges assigned very high values to human life. Over the five years, they valued the 130 decedents at numbers ranging from 220,000 yen to 220 million yen. They valued them at a mean 72 million yen, and a median 66 million (see table 2.6). The number almost exactly tracks the studies of American courts: by one recent commercial jury verdict reporter study, a median of $608,000. 5 The Japanese judges in the published opinion database assigned the very lowest

Table 2.6 Court awards in fatal traffi c accidents A. Selected values (litigated opinion database) Min Male Age Age ≥65 years Litigated value Award premium

0 220 −1.510

25th

Mean

63.1% 38.5 16.9% 52,200 71,900 −.040 .076 20

Median

34

75th

Max

57.5

65,900

82,500 .140

86 220,000

.295

.761

B. Litigated value of life and formula-based (using mean wages) value 1. Regression Litigated value = 2,835,451 + 1.180 Formula value (9.05) 2. Correlation coefficients | Litigated value Formula value Award premium -----------------+-------------------------------------------------------------------Litigated value | 1.0000 | Formula value | 0.6204 1.0000 | 0.0000 | Award Premium | 0.6284 0.0104 1.0000 | 0.0000 0.9052 C. Value and age Litigated value of life = 51,100,000 + −22,419 Age2 + 1,497,177 Age + 9,658,173 Male 4.74 3.66 1.86 Maximum of litigated value occurs at age 33.4 years Maximum of formula-based value occurs at age 34.1 years Source: Daiichi hoki shuppan 2013; search of opinions involving fatal traffic accidents, published 2005–9. Notes: n = 130. Award premium equals the litigated value of life less the formula-based value (Formula value), divided by the litigated value.

24

Chapter Two

value (220,000 yen) to a male of undisclosed age, but the next ten values ranged from 18 to 29 million yen. Of these ten, six were male and four female. One was aged 52 years, three were children, and the other six were aged 70 years or older. Judges assigned the highest value to a young nineteen-year- old man (220 million yen). They valued all the top ten decedents at 117 million yen or more. Seven of the ten were male and three female. Four were minors, and the other six ranged in age from 22 to 59 years. One had worked as a corporate director, and one as vice president to a securities fi rm. According to another source (Songai hoken 2012, 118), the two highest Japanese wrongful- death recoveries to date both went to the heirs of male physicians: one for 529 million yen for a forty- one-year- old ophthalmologist, and the second for 368 million yen for a thirty- eightyear- old physician. (b) Compared to Formulaic Values Compare these observed values with the numbers that one would obtain if one applied the stated formula (section III.C.2, above) using mean wages. Suppose one used the sex and age of each of the decedents in the published opinions and calculated the value of his or her life by presentvaluing mean wages and adding the standard amounts for pain and suffering and funeral expenses. Figure 2.1 graphically illustrates the resulting relationship between the actual awards and the predicted awards (using the standard formula). The figure gives the value that the courts ascribed to the lives of the victims on the vertical axis. It gives the value that would follow from using mean wages and standard pain and suffering amounts on the horizontal. The figure suggests two results. First, the litigated and formula-based values are strongly and positively correlated (see correlation coefficients in table 2.6, panel B): the values calculated by mean wages closely predict the amounts that the courts actually award. Second, the litigated values are slightly higher than the formula-based values: on average, those heirs who sued in court recovered more than predicted by mean wages and standard pain and suffering figures. The simple regression in table 2.6, panel B, makes the point more precisely. The litigated values of human life are 1.18 times the predicted val-

A Tort System That Works

25

Figure 2.1. Litigated awards and formulaic awards. Value ascribed by the court to human life is on the vertical axis, and the formula-based value using national mean wages on the horizontal. Source: Daiichi hoki shuppan (2013), Hanrei taikei database, search of opinions involving fatal traffic accidents, published 2005– 9.

ues, plus a flat 2.8 million yen. What is more, this fractional premium (“Award premium” in table 2.6, panel B) to the litigated value is not a constant 18 percent. Instead, the fractional premium itself rises with the observed values (see table 2.6, section II.B correlation coefficients). (c) Value and Age Recoveries for victims of fatal accidents peak in midcareer. This follows from the fact that earnings peak in midcareer, and courts value lives by present-valuing earnings. Figure 2.2 illustrates this phenomenon graphically: the observations initially rise with age and then fall. In table 2.6, panel B, a simple regression again makes the point more precisely. According to this regression, the litigated value of human life peaks at age 33. As one would expect, the same point holds true for the formulaic value of life calculated by mean wages. The analogous regression on those predicted values peaks at age 34.

26

Chapter Two

Figure 2.2. Court awards by decedent age. Vertical axis gives the value ascribed to the decedent’s life by the court; the horizontal axis gives the decedent’s age. Source: Daiichi hoki shuppan (2013), Hanrei taikei database, search of opinions involving fatal traffic accidents, published 2005– 9.

(d) Selection into Litigation That the observed values for human life lie above the amounts calculated through the judicial formula need not mean judges ignore the formula. More probably, it reflects the way heirs of victims with plausibly high claims tend to select into litigation. In part, plaintiffs with high claims litigate for the well-known reason that the litigation process selects for high-stakes claims generally: only parties with a major dispute will fi nd it worthwhile to pay the attorneys fees involved in litigation (Landes 1971; Posner 1973). And in part the heirs to victims with plausibly high claims select into litigation because they have inside information about the decedent’s high earning capacity. Parties settle the vast majority of disputes out of court. They settle them because they agree about what a court will likely do. Given the costs (mostly attorneys fees) of litigation, they both do better by avoiding trial and settling “in the shadow” of that expected litigated outcome.

A Tort System That Works

27

Given this mutual incentive to settle, parties tend to litigate only those disputes where they are both optimistic (Landes 1971; Posner 1973). They fi nd it hardest to settle when they cannot agree about what a court will do, in other words, and cannot agree because they each think they will do better than their opponent thinks they will. Parties tend not to be mutually optimistic by large margins when the relevant information is mostly public. Suppose a case turns only on the interpretation of a contract. If both parties have a copy of the contract, they will both hire a lawyer. For the most part, those lawyers will give similar advice. The parties will estimate the likely outcome in court in similar terms, and cut an out- of- court deal. Japanese heirs and insurers will settle their disputes out of court when they can. After all, they have no more interest in paying attorneys than heirs and insurers anywhere else. If a truck driver hit a seventy-two-year- old grandfather, damages will not be an issue. Given that the man was retired, the courts will not award lost earnings. Given that no one depended on him for his salary (which, being retired, he did not earn), they will award pain and suffering in the 20– 25 million yen range. If a truck hit a forty-year- old securities analyst, the parties will fi nd settlement harder. The man earned performance-based pay. In bad months, he collected modest amounts. But when his analysis worked, he earned his fi rm large profits, which paid him enormous bonuses. Necessarily, the analyst’s heirs will have a good sense about how successfully they will be able to persuade a court about his high future earnings; by contrast, the truck driver’s insurer will only be able to guess at their likely success. Necessarily, disputes about forty-year- old securities analysts are more likely to end in litigation than disputes over seventy-twoyear- old grandfathers. Hence the bias that mutual optimism introduces into the population of disputes that end in trial: disproportionately, they will be cases where the plaintiffs have favorable inside information about the decedent’s earning capacity. No one will litigate offsetting cases where the defendants have negative inside information about the decedent’s earnings— traffic accidents being what they are, the driver who hits a pedestrian does not have inside information about the pedestrian’s future earnings. But the resulting bias follows straightforwardly: the cases that end in trial will be cases where the plaintiff believes he can prove unusually high wages.

Chapter Two

28

4. liability insurer payments (a) Insurance Coverage Under Japanese law, owners must buy liability insurance for their cars. The amount they must maintain has increased over the years. Since 1991, they must own policies paying at least 1.2 million yen per injured victim and 30 million yen per death (Songai hoken 2012, 49; Ariyoshi et al. 2012, 212). Most Japanese owners add extra coverage. As of 2012, 73.1 percent of Japanese drivers held liability contracts paying additional amounts for injury and death (Jidosha hoken todofuken 2012). In almost all (99.2 percent) these contracts, this extra coverage placed no cap on the amounts payable for wrongful death (Naikakufu 2012, 81). For the majority of fatal traffic accidents, however, the 30 million yen cap itself will not operate as a constraint. After all, for a majority of those who die in traffic accidents, 30 million yen covers the court-recognized value of their lives. The reason for this turns on the age composition of the victims. Consider table 2.7: the age of the people who died in the accidents. The fraction of accident victims age 65 or over rises steadily over time, and by 2011 had reached 56.8 percent.6 In court, the heirs of these victims would receive 1.5 million yen for funeral expenses, pain and suffering, and the victim’s present-valued net expected earnings. For an elderly man or woman no longer supporting children, courts will award

Table 2.7 Age composition of fatal accident victims 1

2

3

4

5

6

7

8

Age, years

2011

Total

Bicycle

Pedestrian

Male

2006

2001

1996

≤15 16–24 25–29 30–39 40–49 50–59 60–64 ≥65

2.4 8.4 2.1 5.6 6.9 10.9 6.9 56.8

134 507

36 53

58 47

66.3 89.5

2,071a

261

546

87.0

2,738

506

1,319

71.8

3.6 10.0 2.9 7.2 5.9 11.4 8.3 50.8

5.6 13.6 4.8 5.1 6.5 12.8 8.4 43.3

3.3 21.4 6.1 7.4 10.2 12.4 7.6 31.6

Source: Naikakufu (various years). Notes: Columns 1 and 6–8 give the percentage of fatal traffic accident victims in that year among various age groups. Columns 2–4 give the total numbers of 2011 victims and the numbers who died on bicycles or while walking (Pedestrian). Column 5 gives the fraction of males among the age groups for 2011. Note that columns 2–5 sum the number of victims within the 25–64 age range. a Total number of all deaths, from age 25 to 64.

A Tort System That Works

29

20– 25 million in pain and suffering. Provided he or she has retired, they will award no net expected earnings. Necessarily, those amounts will total less than 30 million yen. Unfortunately for the decedent’s heirs, the mandatory policies impose other limits. By regulatory order, they pay only 600,000 yen for funeral expenses, for example, and 11 million for pain and suffering.7 For amounts beyond those limits, the heirs will need to claim under the optional policies or against the defendant’s own assets. Despite these limits, however, they do offer claimants some offsetting benefits. Rather than require a victim to prove negligence, they switch the burden of proof and entitle victims to collect unless the defendant can prove both the victim’s negligence and his own lack of negligence. 8 They pay lost wages for working-age victims equal to the greater of actual wages and the national mean.9 And they slash benefits for the victim’s negligence only when his comparative negligence (his share of the overall negligence that caused the accident) is at least 70 percent.10 In disputes over traffic accidents in Japan, 43.5 percent of the claimants in 2010 collected compensation only from the mandatory insurance policy (see table 2.8). Apparently, they make do with more limited pain and suffering awards. Do note, however, that the 43.5 percent follows straightforwardly from the fact that most of the 56.8 percent of the victims will be retired—and hence unable to assert lost wages that might otherwise push their claims beyond the 30 million yen cap. More preTable 2.8 Mean amounts paid by liability insurers in fatal traffi c accidents

Year

A Total deaths

2006 2007 2008 2009 2010

7,272 6,639 6,023 5,772 5,745

B Compensated deaths 6,168 6,029 5,482 5,128 4,922

C Mandatory insurance only

D Mean amount in (C)

E Mandatory + optional insurance in (D)

F Mean amount

2,871 2,689 2,218 2,297 2,140

1,761 1,589 1,392 1,513 1,498

3,297 3,340 3,264 2,831 2,782

3,642 3,669 3,636 3,559 3,530

Sources: Naikakufu (2013), 81; Songai hoken (2012), 84. Notes: Total deaths (column A) are as reported by the police, within thirty days of accident (see table 2.3). Compensated deaths (column B) are those in which some amount of liability insurance was paid (see table 2.3). Column C gives the number of cases in which the only amounts paid were from a mandatory insurance policy, and column D gives the mean amounts (in ten thousand yen) paid in column C. These calculations are based on means that include payments from the small number of mutual aid associations excluded from table 2.4. Column E gives the number of cases in which amounts were paid from both mandatory and optional insurance contracts, and Column F gives the mean amounts (in ten thousand yen) paid in column E. Note that B = C + E.

30

Chapter Two

cisely, 5,745 people died in traffic accidents in 2010, and 4,922 (86 percent) received some compensation from an insurer. The 2,140 who recovered only from the mandatory policy received a mean 15.0 million yen. The 2,782 who collected as well from a supplementary policy collected a mean 35.3 million yen. (b) Insurance Payments The heirs of traffic victims take a discount for not suing the driver who killed the victim, and superficially the discount seems enormous. After all, judges in the published opinions valued the lives of accident victims at a mean 71.9 million yen (table 2.6). Insurers paid much less. In 2010, 2,140 claimants collected only from the mandatory insurance policy, and only a mean 15.0 million yen (table 2.8). Another 2,782 claimants collected additional amounts from a supplementary liability policy, for a mean 35.3 million yen. The obvious weighted average comes to 26.5 million yen—36.8 percent of the mean value ascribed to life in the published cases. This seemingly obvious conclusion, however, is wrong: for three reasons, claimants do not take a 63 percent haircut to settle their claims out of court. First (as noted immediately above), the heirs in court are disproportionately the heirs of victims in their peak earning years. Judges award the highest amounts for victims aged 33– 34 (table 2.6), and the median age of the victims in the published option database was exactly 34 (table 2.6). Judges award no lost earnings to people who have retired, and only 16.9 percent of the victims in the published opinion database were 65 or older. By contrast, the majority (56.8 percent, in 2011) of the victims of fatal accidents are over 65 (table 2.7). Most would be retired and have no lost earnings to collect. Their heirs would collect funeral expenses. They would also collect pain and suffering, but, because most of the victims were no longer custodial parents, they would not even collect much pain and suffering. Second, even among the heirs of working-age victims, those who select into litigation will be the heirs of victims with above-average earnings (as explained above). Indeed, they prove values of life that are even 2.8 million yen and 18 percent of wages higher than the values for people earning wages at the national mean for their age. Third, heirs receive in court only the value of the victim’s life net his comparative negligence. In table 2.6, I give the value the courts ascribed

A Tort System That Works

31

to a decedent’s life. This is not the amount judges awarded. Instead, it is the value they assigned before a comparative negligence offset. By contrast, in table 2.8 I give the amount actually paid by the liability insurers. This is not the value that judges assigned to human life. Instead, it is the amount insurers paid after the comparative negligence offset. In the published opinion database, judges offset the plaintiffs’ recovery for the victim’s comparative negligence by a mean 15.07 percent. By regulatory order (as explained above), under the mandatory policies, insurers dock recovery only when the victim was responsible for at least 70 percent of the negligence. Under the optional supplementary policies, the insurers face no such restriction. And in the published opinion database, judges ordered a comparative negligence offset in 47 percent of the cases. Among those cases with at least some victim negligence, they docked recovery by 10 percent in 18 percent of the cases, by 15 percent in 8 percent of the cases, by 20 percent in 16 percent of the cases, and by 25 percent or more in 58 percent of the cases. Consider a different exercise. Suppose I take the age distribution in table 2.7. And suppose I take the mean wages for each age group, and the standard amounts for pain and suffering. With these numbers, suppose I estimate the value that the courts would ascribe to the lives of the decedents with the standard formula. Under this exercise, the sum would be 182 billion yen. The amount actually paid by the insurers in table 2.8 is 130 billion. Hence, the conclusion: heirs are collecting (130 billion/182 billion =) 71.4 percent of the value that a court would assign to the decedent’s life. Suppose the 15 percent comparative negligence offset from published opinion database generalizes. These calculations suggest that heirs are taking a discount of 14 percent for avoiding litigation.

IV. Comparative Negligence Japanese judges did not just routinize the calculation of the value of human life; they also routinized the determination of the driver’s and victim’s relative negligence.11 As the number of traffic accidents rose during the 1960s and 1970s, so did the number of court cases over those accidents. In each, the judges needed to determine the level of any comparative negligence offset. The Civil Code itself mandates that offset but beyond that gives little detail.12

Chapter Two

32

As they examined each other’s opinions, judges began to realize that they were not handling their cases uniformly. Much of the variation was specific to a judge, of course: some judges more readily docked a victim’s recovery than others. Yet some of the variation was systematically local: although Japanese judges are transferred from court to court at about three-year intervals (see chap. 7), cities developed their own distinctive jurisprudence. Faced with this diversity, the judges in the traffic section (Section 27) of the Tokyo District Court decided to systematize the adjudication. Daniel Foote (1995, 28) describes the process: With the standardization of the compensation amounts, the most significant remaining reason for differences among cases was the issue of comparative fault: differing views on the proportion of negligence could have a major impact on the outcome. Again, Section 27 set out to compile standards in this area. First, [two leading judges] prepared charts setting out the most common accident scenarios. They then had all the judges in the Section list their ratings for the degree of fault in the various scenarios. . . . When it turned out that the scores were similar, they refi ned and compiled the standard into a series of charts specifying the degree of fault.

The judges held conferences. They convened study commissions. And by the late 1960s, they could issue a set of detailed guidelines on comparative negligence. Faced with demand from attorneys and insurers for the information, they published the guidelines in 1975. They have since revised them several times and published the latest version in 2004.13 The guidelines typically devote one page per accident. The top half of the page gives a diagram: an intersection, perhaps, with car A entering, and hitting car B coming in the opposite direction and turning in front of A. A table below the diagram specifies the preliminary negligence assignments. Often, it suggests adjustments for speed of travel or turn signals, or the time of day, weather conditions, age of victim, or traffic light status. The guidelines facilitate settlement for a simple reason: both drivers will anticipate the same comparative negligence offset. The approach is profoundly bureaucratic and distinctly second best, of course: the judge ignores much of the individual variation. He does not care that the mother driving car A ran the yellow light because she was frantically distraught. He does not care that her two-year- old was running a

Figure 2.3. Comparative negligence offsets. Car A enters the intersection (cars travel on the left in Japan), while car B enters from the opposite direction and makes a right turn in front of A. The figure provides a base assignment of 20 percent comparative negligence to A, and 80 percent to B. If B had nearly completed the turn by the time A entered the intersection, A’s negligence rises by another 20 percent. If A is traveling fi fteen kilometers over the speed limit, it rises 10 percent, and if thirty kilometers over it rises 20 percent. If B has not slowed in anticipation of the turn, if B turns in front of A at the last minute, or if B has failed to signal his turn, A’s negligence falls 10 percent. Source: Tokyo chisai minji kotsu sosho kenkyu kai, ed., Minji kotsu sosho ni okeru kashitsu sosai ritsu no nintei kijun [Standards for determination of comparative negligence rates in civil traffic litigation], 38 Bessatsu hanrei taimuzu 1, 237 (2014) (5th ed.) (produced by Tokyo District Court). Reprinted with the permission of Tokyo District Court Civil Traffic Accident Litigation Research Committee (Tokyo chisai minji kotsu sosho kenkyukai) (holder of the copyright), and the K.K. Hanrei Times Sha.

34

Chapter Two

104- degree fever and screaming incessantly. He does not care that car B anticipated the green light because the driver was late to a vital committee meeting. He simply examines the diagram in the book and applies the numbers. Crucially, however, the approach saves the parties attorneys fees. The drivers could consult an attorney, but why? Buy a “traffic accident for dummies” equivalent and check the diagrams. If the drivers do, they will learn how a judge will likely handle their dispute. Once they know, they have little reason to ask him. Cheaper instead to settle by reference to the expected outcome and pocket the money they would otherwise pay their lawyers.14

V. Conclusions Japanese judges handle traffic accidents differently than do American judges. And they handle them differently in a way reflects a basic contrast between the two legal systems. It is a difference that matters. Rhetorically, it is the difference between envisioning courts as a place to promote justice on the one hand; and seeing them as a forum for resolving fights over nontrivial amounts of money on the other. Mechanically, it is the difference between focusing on the individual and the peculiar on the one hand; and following uniform bureaucratic standards on the other. Analytically, it is the difference between aspiring to the fi rst best on the one hand; and making do with the second best on the other.

Chapter Three

A System with Few Claims Products Liability

C

hapter 2 describes a legal system that works. If a car hits a pedestrian, the victim fi les a claim with the driver’s insurer. They negotiate. In time, they settle. They settle, however, in “the shadow of the law.” They settle by reference to what a judge would order the driver to pay the pedestrian, if they bothered to litigate the case. Given that litigation is expensive, they do not bother. Instead, the driver’s insurer pays the pedestrian the amount a judge would order him to pay, and everyone pockets the fees he would otherwise pay his lawyer. Chapter 2 describes judges who are usually generous. When a pedestrian and driver do litigate a quarrel, judges award relatively large amounts. The other parties then settle in the shadow of these cases—for amounts that reflect a price for human life at least as high as the price an American judge would assign, and probably higher. Japanese judges produce these decisions that parties can anticipate because they aim for the “usually right” second best. Not attempting any heroic fi rst best, they do not try to match their awards to the idiosyncrasies of the particular case. Instead, they apply simple, clear, and uniform rules. The result is a system that is both predictable and generous. As the data from this chapter and chapters 4 and 5 indicate, it is also quite fast. Because it is predictable, the parties settle out of court for amounts that track the judgments judges would award. Because it is generous, they settle for high amounts. And because it is fast, they settle expeditiously. Turn now to an apparently very different tort: product liability disputes. Americans fi le these claims massively. They litigate intensely. And as they do, fi rms fail, industries stagnate, and R&D suffers. Japa-

36

Chapter Three

nese fi le these claims seldom. They litigate them even less. And firms and industries never think to complain. Why the contrast? Do American fi rms sell more dangerous products? Do Japanese fi rms escape compensating the injuries they cause? Or do American fi rms settle claims for damages they never caused? Does the Japanese legal system block valid claims? Or does the American system reward fraudulent ones? The reason for the cross-national contrast again lies in the distinction between second- and fi rst-best approaches. Most modern products are safe—for the simple reason that most modern Japanese and American consumers are rich; that most rich people willingly pay for safe products; and that manufacturers do best in competitive markets when they offer consumers what they want. Japanese judges with a second-best approach apparently understand this. They apparently realize they cannot reliably distinguish the rare defective product from the putative defective product advanced by the fraudulent attorney. Rather than try more than halfheartedly, they treat all product liability claims with skepticism. By contrast, Americans have built a legal system around a fi rst-best obsession with giving every claimant a plausible shot at recovery. Alas, American judges are fallible humans. Predictably given their very ordinary limitations, they have not created a truly fi rst-best regime. Instead, they have produced courts where fraudulent attorneys sometimes extort massive amounts from risk-averse manufacturers. I begin by briefly (this is not a book on the United States, after all) outlining the course of products liability law in the United States (section I). I then turn to the law in Japan (section II). I describe suing and claiming behavior (section III), and explore why the numbers are low in Japan (section IV). I conclude by asking why those numbers are so high in the United States (section V).

I. Products Liability in the United States A. Tobacco In the 1990s, Mississippi attorney Richard “Dickie” Scruggs sued four cigarette companies on behalf of forty-six states.1 By tricking consumers into believing that smoking was safe, the fi rms had increased the number of people who smoked. By killing them through smoking-related dis-

A System with Few Claims

37

eases, they had inflated the states’ Medicaid bills. For that cost, they now owed the states compensation. It was a bizarre claim in any of several ways. It was bizarre to claim anyone in the last half century could have thought smoking safe. The news about cancer and heart disease had been clear and relentless for decades. Indeed, concluded Kip Viscusi (1999, 584), “on average people tend to overestimate the risks of smoking.” It was bizarre to claim smoking could have increased government health care and pension costs. By augmenting excise tax revenue and killing its victims quickly, smoking generated for the state and federal governments a net gain. Smoking “increases costs associated with health care,” explained Viscusi (1999, 577), “but it also decreases nursing home and pension costs because smoking leads to earlier mortality.” It was bizarre (to put it charitably) the way Scruggs collected the evidence against the tobacco fi rms. He located a friendly paralegal at a law fi rm that represented one of the tobacco fi rms who had taken fi rm documents with him when he left. Scruggs and his partners cosigned on the paralegal’s loans to buy two cars, bought him a house, and took the documents. 2 The tobacco fi rms had sold a legal but dangerous drug to buyers who knew it was hazardous and saved the state governments net outlays—but never mind. They caved. They agreed to slash marketing efforts. They agreed to cut lobbying activities. They agreed to pay the states $246 billion over twenty-five years. And the intellectual class cheered. The New York Times lamented only that the terms were not harsher still. 3 Scruggs’s law fi rm’s take: $900 million.4 B. Asbestos Scruggs had made his fortune suing the asbestos fi rms. The litigation had a major effect on the American economy. The Towers Perrin Tillinghast (2004, 2– 3) consulting fi rm estimated the total cost of the tort system in 2003 at $246 billion and attributed $9 billion of that amount to asbestos. In many ways, the material had seemed the ideal insulator: it did not burn, did not conduct electricity, absorbed sound, and did not react with most chemicals. But it could also injure and sometimes kill, particularly among people who smoked. Over the 1965– 2009 period, it apparently killed between forty thousand to over three hundred thousand (Carroll et al. 2005, xix, 15–18).

38

Chapter Three

The tort law behind Scruggs’s asbestos campaign was as wrong as the law behind his tobacco litigation. The plaintiffs had mostly faced asbestos at work—in a fundamentally contractual setting. After all, employers and employees negotiate their relationship in a competitive market. Employers pick employees who offer the right characteristics (talent, work habits, and expertise) at the best price (salary, insurance, and other benefits). Employees pick employers who offer the best combination of pay, surroundings, and other job characteristics. If a job presents a health risk, employees take it only when the fi rm offers compensation and other characteristics that offset that risk. Scruggs’s smoking victims had bought cigarettes for the pleasure, knowing the risk they took; his asbestos plaintiffs had taken the jobs for the wages, knowing the risk they took. Nothing about the asbestos risk was new news. In 1898, the founder of the Johns-Mansville asbestos giant had died of “dust phthisis pneumonitis.”5 By 1918, insurance companies were refusing asbestos workers life insurance.6 And the very term “asbestosis” dates from 1925 (Bartrip 2004). Information about the risks was there to be had. The workers took the job for the money and sued when the risks came true. Nonetheless, the asbestos litigation began in the 1970s and grew exponentially. Rather than limit their clients to people with signs of disease, lawyers like Scruggs apparently recruited potential clients and paid for X-rays and asbestosis screenings (such as they were) themselves (Carter 2008). By 2002, 730,000 plaintiffs had sued 8,400 fi rms on asbestosrelated claims. The fi rms had paid their lawyers $21 billion, the plaintiffs’ attorneys $19 billion, and the plaintiffs themselves another $30 billion (Carroll et al. 2005, xxv–xxvii, 6, 70– 71, 79, 109, chap. 5; White 2004)—and 80 fi rms had fi led for bankruptcy. The attorneys fi led most of the suits in a very few places. As of the mid1990s, they fi led a quarter of the new state- court suits in three Texas counties. Within a few years, they shifted new suits to two Mississippi counties. From 1998 to 2000, they fi led almost 20,000 asbestos suits (10 percent of all asbestos cases) in these two Mississippi jurisdictions (Carroll et al. 2005). In 2000 alone, they fi led suits on behalf of 3,000 plaintiffs in Scruggs’s favorite courts in Jefferson County, Mississippi (pop. 9,740) (Boyer 2008). They did not pick these courts by roulette. By bringing a case in a favored county, calculates economist Michelle White, a plaintiff increased his recovery by an amount between $1.7 and $2.6 million (White 2004). From asbestos, Scruggs’s law fi rms’ take: $25 million.

A System with Few Claims

39

In the mid-2000s, the plaintiffs’ attorneys shifted their focus from asbestos to silicosis—but sometimes with the same plaintiffs. In 2005, Texas federal judge Janis Jack was handling a massive number of silicosis suits when she noticed that twelve doctors had diagnosed 9,083 of the plaintiffs. Hired by lawyers and screening companies, these twelve had identified a disease eight thousand other physicians had not.7 The X-rays for silicosis and asbestosis are very different and very few patients have both diseases. Yet Jack’s plaintiffs did:8 When Dr. Harron fi rst examined 1,807 Plaintiffs’ X-rays for asbestos litigation (virtually all done prior to 2000, when mass silica litigation was just a gleam in a lawyer’s eye), he found them all to be consistent only with asbestosis and not with silicosis. But upon re- examining these 1,807 MDL Plaintiffs’ X-rays for silica litigation, Dr. Harron found evidence of silicosis in every case.

All except one of the ten thousand cases, Jack concluded, stemmed from “fatally” flawed diagnoses “manufactured for money” (Behrens and Schaecher 2010, 1). Yet even when judges identify attorney fraud, they do not necessarily punish. For masterminding and carrying out the scam, she fi ned the kingpin law fi rm $8,250. From asbestos and tobacco, Scruggs himself moved to insurance: to suing State Farm over its Hurricane Katrina payments. State Farm had treated claimants unfairly, he argued. For 640 claimants, he obtained $80 million. He also negotiated $26.5 million in attorney fees, but the litigation would be his fi nal undoing. To convince State Farm to pay, he donated lavishly to the state attorney general’s election campaign and convinced him to launch a criminal investigation against State Farm. But the money was too good. When he fought with the other attorneys over the fee, he tried to bribe a judge. The judge reported him to the FBI, and prosecutors fi led charges. Scruggs found himself disbarred and sentenced to five years in prison, and seven years in prison for trying to bribe a second judge (Parloff 2008; Lange 2008). C. Small Planes Tobacco and asbestos may be the most notorious fields in the world of American products liability litigation, but the damage to more mundane industries has been just as severe. Many industries it has left dev-

40

Chapter Three

astated. It has raised costs, hiked prices, and slashed revenues. It has cut research. It has chased enterprises from the field. Take the small plane industry. Pilot error causes almost all accidents (75–100 percent) in these planes, but “90% of all accidents involving a fatality or serious injury result in a lawsuit against aircraft manufacturers” (Martin 1991, 493; Nelson and Drews 2008, 426, 428). To cover the cost of the expected litigation, manufacturers hiked the price of new planes. In 1977, liability expenses had come to $1,420 per plane. By 1986, they had soared to $75,000–$80,000 (Nelson and Drews 2008, 428). As the price of new planes rose, consumers made do with older planes or built their own. In 1978, manufacturers sold 17,811 new planes. In 1994, they sold only 928. Yet from 1971 to 2000 the number of “amateurbuilt aircraft” climbed 674 percent (Nelson and Drew 2008, 427– 28). Older planes are not as safe as new ones, and homemade planes are not as safe as those professionally built. Randy Nelson and James Drews (2008, 436) conclude that the law straightforwardly lowered product safety: “The decrease in new plane shipments resulted in an additional 22,534 accidents and 7,887 fatalities over the period 1981– 2005.” In time, the industry recovered—but not on its own. It recovered only because Congress intervened. In 1994, it enacted the General Aviation Revitalization Act to limit claims against the airplane manufacturers.9 With the new limits in place, the industry began to rebound. D. Pharmaceuticals A similar phenomenon occurred in the pharmaceutical industry. Among childhood vaccines, one of the most effective and safe has been DPT (diphtheria, pertussis, tetanus). Nevertheless, during the 1980s, the plaintiffs’ bar began to sue the manufacturers of the vaccine for purported side effects. The “diphtheria and tetanus components of the DPT vaccine are generally considered very safe,” wrote Richard L. Manning (1994, 257n14). But the pertussis component was “suspected of carrying a small risk of very serious side effects.” Only suspected—physicians disagreed about whether the “severe reactions associated with the pertussis vaccine [were] causal in nature or merely coincidental.” Given the enormous cost of litigation, however, the price of the DPT vaccine soared anyway. From under $10 in 1980, it soared to over $150 by 1987 (Helland and Tabarrok 2006, 16). Some fi rms left the market. Some children did without.

A System with Few Claims

41

The phenomenon repeated itself with several other vaccines. Because of litigation costs, concluded one scholar, the “wholesale price of the oral polio vaccine is estimated to be more than 300 percent higher than it would have been under traditional negligence rules.” Because of product liability, “the measles, mumps, and rubella vaccine prices are on the order of 50 percent higher than they otherwise would have been” (Manning 1994, 273). Prices rose. Again, some children did without. Probably the most tragic result involved Bendectin. Merrell Dow had introduced the drug in 1956 to ease morning sickness. Pregnancy routinely causes mild discomfort, but with 0.3– 2 percent of new mothers it can cause severe nausea and vomiting (Eliakim, Abulafia, and Sherer 2000). In extreme cases, it can kill—as it killed Charlotte Bronte in 1855. In the early 1980s, plaintiffs’ attorney Melvin Belli masterminded a massive litigation campaign against Bendectin’s manufacturer, Merrell Dow. Bendectin, he argued, caused birth defects. In truth, Belli had no evidence that it injured anyone at all. Many women have morning sickness, many women took Bendectin, and babies occasionally have birth defects. Predictably, given these odds, some women who took Bendectin gave birth to babies with birth defects. As a witness, Belli had a physician who had written an article asserting that Bendectin caused these defects. Alas, the doctor had faked his data (Huber 1993, 126). Evidence that Bendectin caused harm simply did not exist. Notwithstanding science, litigation costs soared ($100 million) and Merrell Dow pulled Bendectin from the market. Women with morning sickness now have nothing comparable to which to turn. Worse, pharmaceutical fi rms have stopped almost all research in contraceptive or pregnancy-related drugs (Viscusi and Moore 1993, 162). As several public-health scholars put it, “no drug with an obstetric indication has been approved since 1995” (Wing 2010, 825, 828): [Despite a] record of safety that speaks to no measurable reproductive risks to the mother or the developing fetus, a number of nonmeritorious lawsuits alleging causality in fetal malformations were fi led and went to trial. Because of the relatively high frequency with which nausea and vomiting of pregnancy are encountered, the risks of nonmeritorious litigation and its costs were prohibitive. The manufacturer of Bendectin subsequently went out of business as a result of the costs of litigation defense, leading other drug developers to be wary of drug development for obstetric indications.

42

Chapter Three

Eventually the pharmaceutical industry recovered, too—but again not on its own. Because of the crisis in the industry, Congress intervened. It passed the National Childhood Vaccine Injury Compensation Act.10 Some courts in what had been proplaintiff states began to scale back the reach of their earlier opinions.11 Again, in time the industry began to rebound.

II. Products Liability in Japan A. Legal Doctrine 1. in modern democracies Doctrinally, Japanese products liability law differs very little from that in the United States. Indeed, it differs very little from the law in any of the wealthy modern democracies. The “blackletter product liability law among the jurisdictions” of modern democracies, concluded comparative law scholar Mathias Reimann (2003, 761), “points to a considerable uniformity.” On the substantive law, wealthy democratic regimes do not much diverge. Instead, they show “a growing agreement about such basics as the paradigm of strict manufacturer liability.” According to Reimann (2003, 802, 788, 793), the wealthy democratic regimes enforce demanding standards. The “United States is no longer the only country with tough product liability rules,” he observes. “Almost all industrialized nations have them today.” For this “tough” law, most countries (including Japan, but not the United States) detail the doctrine in a special statute. Most also let plaintiffs bring their claims about defective products under generic tort (e.g., the fi rm negligently failed to inspect the product) or contract law (e.g., the firm promised a safe product but failed to perform) besides. The “tough” doctrines that rich countries impose resemble each other down to the detail. Most use a strict liability standard (Reimann 2003, 775): “The clear majority of special product liability regimes today do not require the plaintiff to show that the defendant was at fault.”12 Rather than require a victim to show that the manufacturer was “negligent,” in other words, they require only that he show that the manufacturer sold a defective product that caused the accident.13 For manufacturing defects, continued Reimann (2003, 775, 776), the strict liability is actually strict: “in the vast majority of regimes, manu-

A System with Few Claims

43

facturing defects trigger liability regardless of due care.” If a manufacturer sells a product that breaks because of a bad weld, the plaintiff need not show how the bad weld happened. He need only show that the weld was bad and caused the injury. For some other products liability claims, however, the strictness is less substantive than nominal: “When it comes to the liability of manufacturers for design defects and insufficient warnings, strict liability is often tempered with elements normally associated with negligence.” If a manufacturer is “strictly liable” for selling defectively designed products, for example, liability may turn on whether a design was in fact defective. Suppose a judge insists that a design is defective only if a plaintiff can show that the manufacturer invested insufficient care in designing or testing it. Realistically, by requiring that he prove “insufficient care,” the judge can transform the test from strict liability into negligence. 2. in japan Reimann described products liability law in wealthy democracies generally, but his description applies nicely to Japan specifically. In 1994, the Japanese legislature passed a statute (effective 1995) holding manufacturers strictly liable for injuries caused by their products.14 It modeled the law on the European Community Directive of 1985,15 and at its heart lies Section 3: A manufacturer must pay compensation for any damages to the life, health, or property of another person, if those damages were caused by a defect in a product that it manufactured, processed, or imported. . . .

The rule applies to all products sold after July 1, 1995.16 As with its European antecedent, it does not apply to services. Even without the 1995 statute, Japanese injured by defective products could sue. Prior to the statute, they could bring products liability claims under the standard rules of tort or contract. Since 1995 they can still sue under tort or contract, but they can add claims under the new products liability statute to boot. Basic Japanese tort law dates to the 1896 Civil Code.17 Modeled after the German equivalent, the code mandates a negligence regime for torts—precisely the regime that applies to the automobile accidents in chapter 2 (Sec. 709):

Chapter Three

44

Sec. 709: He who infringes a right of another, either intentionally or negligently, must compensate that person for any damages he causes.

To determine whether a defendant has acted negligently, the courts often (not always) adopt a balancing test close to Judge Learned Hand’s famous formula in Carroll Towing.18 Possibly it serves to bring [the rule] into relief to state it in algebraic terms: if the probability [of the accident] be called P, the injury, L; and the burden [of preventing the accident] B; liability depends upon whether B is less than L multiplied by P; i.e., whether B < PL.

Due care is the care that “a person of ordinary care would show under the circumstances,” explained the Japanese Supreme Court,19 and those circumstances typically require people to weigh risks (Hand’s P), harms (L), and precautionary costs (B). In fact, in products liability, negligence under the Civil Code could sometimes approach strict liability anyway. Take a classic 1994 dispute over a Panasonic television set that caught fi re. The Osaka District Court went out of its way to insist that it did not hold Panasonic to strict liability. All the same, it declared that the fi rm had a duty to sell television sets that were safe for normal use; that a television that catches fi re is not safe; and that—given the fi re—it would “presume” that Panasonic had produced the set negligently. Panasonic could avoid liability only if it proved that it had not been negligent. Unable to rebut the presumption, Panasonic lost. 20 Conversely, strict liability under the new statute can sometimes resemble negligence. As noted earlier, the point is particularly true for design defects. Whether a design is “defective” under Section 3 of the statute turns on reasonable levels of investment in design and testing, as well as the product’s reasonably intended use. “Reasonableness,” however, introduces a cost-benefit analysis, and cost-benefit analysis takes us back to Hand’s negligence standard in Carroll Towing (Tsusho sangyo sho 1994, 96). Before 1995, plaintiffs could often sue in contract. They still can. Japanese judges sometimes declare safety an “implied” contract term. The seller agreed to provide a given product; the deal included an “implied” agreement that the product would be safe; the product was not safe; the

A System with Few Claims

45

lack of safety caused the damage; ergo, the seller was liable under contract for breaching an implied term in the agreement. Take for example a family that bought a toy archery set. When the suction cup at the end of one of the arrows came off, a daughter blinded her brother. The family collected from the retailer, the retailer collected from distributor, the distributor collected from the wholesaler, and the wholesaler sued the manufacturer for indemnity. The manufacturer had an obligation to sell a safe archery set, the judge reasoned. Because the set—in retrospect—had not been safe, the manufacturer was liable. 21 Sometimes judges just duck fi ne distinctions entirely. The strict liability rule in the new statute may require a quasi-negligence analysis. The negligence rule in the Civil Code may approach strict liability. And as scholars in “law and economics” have noted for decades, the two regimes have similar efficiency characteristics anyway. Given the interpretive overlap and the crude similarity between the two, Japanese judges sometimes just ignore the new statute and stay with what they know. Rather than learn the new rules, in other words, modern judges may just apply the law they learned in school. When a plaintiff sued a cosmetics fi rm under both general tort rules and products liability, the district judge held the fi rm liable in negligence without worrying about the products liability statute. 22 When a defendant sold a plaintiff contaminated dumplings, the plaintiff sued in contract and products liability. 23 The district judge held the defendant liable in contract without bothering to apply the products liability statute. And when a plaintiff sued Fuji Heavy Industries (Subaru) over an engine- compartment fi re, the district judge held the fi rm liable without even distinguishing the different doctrines. 24 B. Products Safety In the modern United States and Japan, defective products seriously injure few people. Take figure 3.1: the number of nontraffic accidental deaths in Japan from 1950 to 2010 (left axis), along with the rates per hundred thousand population (right axis). They show a steady decline over the four decades from 1950 to about 1990, and a sharp increase in the two decades since. The reasons for the initial fall in the number of deaths and its later rise are easy to fi nd. As in the United States, so, too, in Japan: the demand for safety increases with income. During the fi rst decades after

46

Chapter Three

Figure 3.1. Nontraffic accidental deaths (numbers and rates per hundred thousand). Number of deaths on left axis, and rates per hundred thousand on the right axis. Source: Kosei rodo sho (various years).

the war, per capita Japanese income climbed steadily and rapidly. Increasingly, consumers demanded and obtained safer products and safer homes. As they did, the rate of nontraffic-related accidents declined. From 30.4 deaths per hundred thousand population in 1950, the rate decreased to 11.9 deaths per hundred thousand in 1987. With the increase in wealth, however, came a demographic shift. As Japanese grew richer, they lived longer and bore fewer children. The median age began to climb sharply, and with it the rate of accidental deaths. By 2010, the rates had returned to those of 1954. Japanese live long today because scientists have eliminated the diseases that killed people earlier in life a half century ago. In 1943, tuberculosis killed 170,000 Japanese (Ikeda, Nadaoka, and Kurashina 2003). Thanks to pharmacological progress, the disease now kills only 2,000– 3,000 people a year. One could tell similar stories about most of the other diseases that assaulted the Japanese population in the midtwentieth century. Notwithstanding modern medicine, though, everyone eventually dies —and in a world where medicine has tamed killers like tuberculosis, accidents have become like cancer, strokes, and heart disease: they kill those who weather everything else that nature sends their way. A Japanese woman may have survived the American bombing raids in the mid1940s. She lived through the postwar food shortage. She avoided tuber-

A System with Few Claims

47

culosis, cholera, and dysentery. Now in her late eighties, she will slip in the kitchen and fall. She will break her hip, and after a few weeks in the hospital will die—of the 28,000 nontraffic accidental deaths in 2004, 6,412 involved falls. She may slip in the bathroom and drown in the tub— 5,584 of the accidental deaths involved drowning. Or she may choke on her food—8,645 of the deaths involved suffocation (Kosei rodo sho 2004, table 2). Overwhelmingly, these accidental deaths kill the elderly. Of the 28,000 accidental deaths in 2004, 42.1 percent involved men and women aged 80 or more, and 79.7 percent involved those aged 60 or older. Of the 6,412 falls, 45.2 percent were suffered by people aged 80 or older, and 81.1 percent by people aged 60 or older. Of the 5,584 drownings, 32.6 percent of the victims were aged 80 or older, and 77.5 percent aged 60 or above. Of the 8,645 asphyxiations, 53.6 percent were aged 80 or beyond, and 88.6 percent were aged 60 or above (Kosei rodo sho 2004, table 2). Few of these accidental deaths over the last two decades, as depicted in figure 3.2, resulted from defective products. Although the numbers will indeed include any death caused by a product defect, they include much more. When an elderly woman falls and breaks her hip, she does not fall because the stool she climbed or the chair she sat in was defective; she falls because she is old and unsteady. She does not choke on a dumpling because it was defective; she chokes because her reflexes no longer work as they once did. To explore the impact of (a) the increase in the fraction of population that is elderly and (b) the decline in the lethality of the traditional killers, I add two curves: nontraffic accidental deaths among those aged 65 or younger, and the accident rate per 100,000 population among this group. Note that the curve no longer turns up during the 1990s: accidental deaths among the nonelderly are not increasing in Japan. Instead, they are continuing their slow and steady fall (with an idiosyncratic jump in the mid-1990s). Note, too, that there is no sharp drop after the mid1990s (instead, the slow and steady decline continues). At least by this evidence, the 1995 strict products liability act did nothing to increase product safety. Overall, Japanese accident rates closely track those in the United States. In 1999, 55,459 people died from nontraffic-related accidents in the United States—a rate per 100,000 of 19.7 (Health Grades 2013). In 2000 in Japan, there were 26,627 nontraffic accidental deaths—a rate per

Chapter Three

48

100,000 of 21.2. The rate is probably a bit higher in Japan because of the older population. More basically, however, both countries are rich, and both have consumers who demand safe products. As a result, both countries have similar accidental death rates. What is more, the accidents in the two countries arise from similar products. In table 3.1, I tabulate the product mix involved in the accidents. I take the data from the reports submitted by a network of participating American hospitals to the National Electronic Injury SurveilTable 3.1 Products generating the most injuries, as reported by hospitals in the United States and Japan A. United States Product

Actual injuries, % 2009

Stairs and railings Beds Desks and tables Bicycles Chairs Basketball Football Bathroom equipment Doors Exercise equipment Playground equipment Toys

20.9 5.62 5.43 4.57 4.63 4.48 4.12 2.99 2.93 2.83 2.44 2.31

B. Japan Actual injuries, % Product Stairs Bicycles Skis Bathroom equipment Roads, paths Chairs Knives Tobacco Automobiles Beds

1994

1995

1996

8.62 6.89 2.00 2.17 2.85 2.00 2.77 2.44 2.46 2.40

7.98 6.32 2.26 2.23 3.10 2.07 2.80 2.78 2.79 2.31

7.73 5.86 3.26 2.39 2.30 2.05 2.66 1.97 3.27 —

Sources: National Electronic Injury Surveillance System (NEISS), ed., NEISS Data Highlights (2009), www.cpsc.gov/library/neiss.html; Kokumin seikatsu sentaa (various years). Notes: For the years listed for Japan, the source reports the ten products generating the most injuries. A dash (“—”) does not indicate 0; rather, it indicates that the product did not fall within the top ten products that year.

A System with Few Claims

49

lance System (NEISS), and by an analogous group of Japanese hospitals reporting to the government. In both countries, a large fraction of serious accidents involve stairs, beds, and chairs—most of the victims were probably old men and women who lost their sense of balance. Similarly, in both countries a large fraction of accidents involve sports: bicycles, basketball, football, and exercise equipment in the United States, and bicycles and skis in Japan. Note that old people do not fall because they slept in a defective bed. Neither do boys hurt themselves in football because they threw a defective ball. C. Doctrine and Safety Although figure 3.1 seems to suggest that the 1995 strict products liability statute did not increase product safety, scholars routinely claim that it did. They argue that manufacturers responded to the law by making products safer. In 1998, for example, Luke Nottage and Yoshitaka Wada (1998, 49; see Nottage 2004, 195– 97) reported that Japanese manufacturers had redoubled their efforts to increase safety: [The new statute] has had an impact on manufacturers, which have responded with a range of measures. There are efforts to improve product safety at the level of the individual fi rm; new committees and guidelines on improving labeling and instructions instituted by industrial associations; more monitoring of accidents by associations. . . . Individual fi rms are expanding their legal section personnel to deal with PL issues.

Several years later, Nottage (2004, 196, 198) again claimed that “manufacturers [in Japan] have become much more conscious of PL issues and accordingly have adopted counter-measures to improve product safety.” In fact, he continued, there is “clear evidence of widespread attempts by manufacturers to improve their product safety and complaint handling procedures, initiated mainly over 1994– 6.” If manufacturers improved safety, no evidence of it appears in accident reports. Take the data collected by the national network of Consumer Centers. Recently restructured as Citizens’ Life Centers, the outlets coordinate the national government’s ostensible efforts to promote product safety. The government began these centers in 1970, and by 1997 had established 330 offices across the country (Kokumin seikatsu sentaa 2010, i; 1988, 4; 1997, 1; 1999, 1; 2008, 6). Given that it sponsored the 1995

50

Chapter Three

statute, if the government had any bias it would cut in favor of identifying an increase in product safety. According to Consumer Center data, the 1995 statute did nothing of the sort. The statute did not cover all purchases; it covered products but excluded services. If it promoted safety, then the number of productrelated accidents should have fallen relative to the number of servicerelated accidents. It did not. Figure 3.2 gives the number of product- and service-related inquiries that the center received. Figure 3.3 then indexes those numbers by their 1987 values. 25 According to figures 3.2 and 3.3, product- and service-related inquiries increased at about the same pace. Both remained stable until the mid-1990s, and then both began to climb steadily (probably because of increased reporting practices). Throughout, the two lines track each other closely. That the law would not have raised product safety is exactly what theory (and common sense) would have suggested. Both the United States and Japan have mostly competitive markets, and both have mostly rich consumers. Product safety is a normal good, the demand for which rises with income. In both countries, consumers will mostly want safe products. Even without strict products liability, the manufacturers that survive in these markets will tend to be the fi rms that offer that safety (and given that the courts had already interpreted the Civil Code to impose rules close to strict liability, the 1995 statute did not radically change the law anyway). To be sure, markets are never perfect. Informational asymmetries,

Figure 3.2. Product- and service-related health and safety inquiries. Source: Kokumin seikatsu sentaa (various years).

A System with Few Claims

51

Figure 3.3. Product- and service-related health and safety inquiries, indexed at 1987 = 100. Source: Kokumin seikatsu sentaa (various years).

negotiation costs, and commitment difficulties can sometimes prevent consumers and manufacturers from contracting for mutually favored levels of product safety. 26 And even rich consumers will not want to pay for zero- defect products. But one can lose the forest for the trees. If rich consumers mostly want safe products, manufacturers will mostly offer them safe products, for a very simple reason: safe products will sell better than unsafe products. For the most part, then, when the Japanese legislature imposed strict products liability in 1995, Japanese manufacturers had little reason to raise safety standards. For the most part, they had been selling mostly safe products anyway.

III. Suing and Claiming in Japan A. Introduction Japanese fi le product liability suits less often than do Americans (section B). They also assert claims less often (section C). They not only fi le fewer suits in court but also fi le fewer claims out of court. B. Suing The contrast in litigating practice between the United States and Japan is massive. 27 In the United States, Mitchell Polinsky and Steven Shavell

Chapter Three

52 Table 3.2 Products liability opinions published (Hanrei taikei search) A. Published opinions

Claim year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2005 2006 2007 2008 2009

No. of claims 1 0 1 3 2 5 5 9 7 4 4 9 2 2 0 2

No. of plaintiff recovery 1 1 2 1 3 4 7 6 2 3 6 1 1 1

B. Summary statistics Amount awarded/amount claimed, where plaintiff recovers 0.4661 Recovery amounts, where plaintiff recovers Low: 41,250 Mean: 27,393,432 High: 177,088,636 Recoveries for wrongful death, where plaintiff recovers Low: 5,500,000 Mean: 60,563,922 High: 177,088,636 Source: District Court opinions recovered under a search for “products liability” (seizobutsu sekinin) in Daiichi hoki shuppan 2013.

estimate that plaintiffs fi le about eighty thousand product liability suits per year. 28 In Japan, hardly anyone sues. To explore how many suits people do fi le, I searched for “products liability” (seizobutsu sekinin) in the Hanrei taikei district court opinion database (as used in chap. 2). The search yielded a set of sixty-three opinions. I report selected summary statistics in table 3.2. Note that the modal number of published opinions on product liability in any given year is 2; the mean is 3.5. Because most court cases in Japan do not generate a published opinion, this search understates the amount of litigation. In 2004, for example, the Hanrei taikei database included 1,358 civil district court opinions of any kind. The district courts actually issued 71,428 opinions that

A System with Few Claims

53

year (44,711, if we exclude default judgments). The parties settled another 51,331 suits before fi nal judgment (Saiko saibansho, Shiho tokei 2004). If court reporters publish 1,358 of the 44,711 nondefault judgments in a year, then 3.5 annual product liability opinions imply a universe of 115 decided cases per year. If plaintiffs fi le 122,759 civil cases, they imply a universe of 316 fi lings per year. Commercial court reporters stay in business by selling subscriptions. Because products liability litigation is so novel, they almost certainly publish a higher fraction of the opinions than in other fields. If so, then the mean number of cases decided annually may be substantially fewer than 115. On its website, the Consumer Center network lists all products liability suits about which it has information. The centers do not just include those cases that yield published opinions. Instead, they list any court case that reaches their attention. All told, they list 142 cases. I report selected summary statistics in table 3.3. The modal number of product liability cases in any given year is 6; the mean is 8.4. The mean time from fi ling to judgment is 2.5 years. Plaintiffs sued and collected on products liability claims long before the 1995 statute. They simply sued in negligence or contract rather than strict liability. Hideyuki Kobayashi estimates that by 1990 the courts had published about 140 products liability opinions. Others place the number of prestatute opinions at 150. 29 Kobayashi writes that most of the cases involved tort (Civil Code, Sec. 709) rather than contract (Sec. 570) law. 30 Given the search results in table 3.2, apparently very few of the early opinions included the phrase “products liability.” Consider then the resulting contrast: Polinsky and Shavell (2010) place the annual number of American products liability suits at eighty thousand; extrapolations from the annual number of published opinions in Japan range from under one hundred to a little over three hundred. The United States has a population of 307 million; Japan has 128 million. Apparently, Americans fi le over a hundred times as many product liability suits per capita as Japanese. C. Claiming Not only do Japanese litigate less frequently than Americans, they also claim less frequently. Out- of- court claims are obviously hard to track. As a crude proxy, however, take evidence from the insurance industry.

Table 3.3 Products liability cases fi led (Consumer Center Data) A. Claims fi led Claim year 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

No. of claims 1 2 6 11 10 10 12 11 12 20 13 7 8 6 6 6 1

Plaintiff recovery

Plaintiff loss

Settlement

0 0 2 2 4 4 6 4 5 10 5 1 2 1 1 0 0

1 0 1 3 5 2 5 3 3 5 6 4 2 2 2 0 0

0 2 3 6 1 4 1 4 4 5 2 1 4 2 2 0 0

B. Summary statistics 1. Litigation Fraction litigated, if fi led: 0.648 Fraction published, if litigated: 0.451 2. Plaintiff recovery Fraction with plaintiff recovery, if litigated: 0.522 Fraction with plaintiff recovery, if published opinion: 0.683 Fraction with plaintiff recovery, if not published: 0.380 3. Amount recovered Amount recovered, if plaintiff successful Low: 30,000 Mean: 19,100,000 High: 117,000,000 Mean recovery, if published opinion: 22,500,000 Mean recovery, if not published: 14,200,000 4. Delays Time to judgment Low: 0.42 years Mean: 2.52 years High: 6.58 years Mean delay, if plaintiff recovers: 2.61 years Mean delay, if plaintiff loses: 2.45 years C. District court, number of cases Tokyo 44 Osaka 22 Nagoya 12 Kobe 6 Kagoshima 5 Sendai 5 Hiroshima 4 Kyoto 4 Nara 4 Yokohama 3 Other 33 Source: Kokumin seikatsu sentaa, district court cases recovered from http://www.kokusen.go.jp/pl_l/index.html.

A System with Few Claims

55

Insurance premiums will reflect the amounts fi rms pay on products liability claims. After all, if a fi rm wants to insure against those claims, it will buy its coverage on a private market. Insurers will quote it rates that depend on industry, coverage, annual sales, and expected liability. 31 The Mitsui- Sumitomo insurance fi rm, for instance, offers products liability coverage to manufacturers, retailers, restaurants, and other service establishments (Mitsui- Sumitomo 2010). Retailers and fi rms in the service sector are not subject to the 1995 act, but they do remain liable under tort and contract law. Mitsui- Sumitomo offers all of them coverage. It excludes claims related to nuclear radiation, asbestos, and intentional or grossly negligent conduct, and covers claims against exported products through separate contracts. The Mitsui-Sumitomo (2010) brochure gives several examples of the prices charged. At the time of the brochure in 2010, a construction firm with annual sales of 500 million yen could buy coverage for 293,000 yen. Under the contract, Mitsui-Sumitomo would pay up to 50 million yen for personal injury per claimant, 300 million yen for total personal injury claims per year, and 30 million yen for total property damage claims per year. In table 3.4, I collect price quotations from a variety of insurers. In the last column, I infer the implied rate per million yen in annual sales. This per-million yen rate starts at 150 yen for 100 million yen total personal injury coverage for a supermarket. It peaks at 2,700 yen for 100 million yen coverage for an automobile repair shop. The Japan Chamber of Commerce (JCC) coordinates products liability insurance coverage for small and medium-sized member fi rms (Nihon shoko kaigi sho 2012). The contracts themselves are sold by commercial insurers. Manufacturers can buy this JCC-mediated coverage if they have three hundred or fewer employees, or three hundred million yen or less in paid-in capital. Wholesalers can join if they have 100 or fewer employees or one hundred million capital or less, retailers if they have fi fty or fewer employees or fi fty million capital or less, and service industries if they have a hundred or fewer employees or fifty million yen capital or less. The Okinawa Chamber of Commerce (Okinawa Chamber of Commerce 2011) details the rates for qualifying small fi rms (it uses the same size cutoffs as the JCC). Table 3.5 gives the rates per million yen in annual sales. These chambers of commerce–mediated rates fall slightly

Chapter Three

56 Table 3.4 Products liability insurance: Commercial carrier rates

Insured Supermarket Household goods Textiles Food retail Carpentry Glass products Plastics manufacturing Toy manufacturing Plastic products Chemical retail Construction Food production Restauranta Restauranta Electronic equipment retail Bakery retail Restaurant Plumber Machine tools Auto repair

Annual sales (¥ million)

Coverage (¥ million)

Annual premium

Implied rates

200 300 150 300 200 300 200 1,000 200 300 500 200 500 300 200

100 500 50 500 100 100 100 100 300 500 300 300 300 300 100

30,050 2 49,860 6 29,100 3 67,010 6 72,740 2 123,500 3 82,990 2 510,000 5 102,120 6 154,610 6 293,0001 192,510 6 501,170 6 333,130 6 231,770 2

150 166 194 223 364 412 415 510 510 515 586 963 1,002 1,110 1,159

300 100 300 200 100

100 300 100 300 100

351,970 4 124,560 6 388,620 4 271,930 6 270,000 5

1,173 1,246 1,295 1,360 2,700

Notes: “Implied rates” are per one million yen in annual sales. “Coverage” is total personal injury claims paid per year. 1 Mitsui-Sumitomo kaijo kasai hoken, K.K. (2010). 2 Kyoei fire, www.kyoeikasai.co.jp/pdf/corp/liability/pl.pdf. 3 Sonpo Japan, www.sompo-japan.co.jp/hinsurance/risk/relief/sh_seizou/index.html. 4 Daido fire, www.daidokasai.co.jp/product/hojin_compensate.php. 5 Tokyo marine, www.tokiomarine-nichido.co.jp/hojin/baiseki/seisanbutsu/index.html. 6 Mitsui-Sumitomo, www.ms-ins.com/houjin/product/biz-protector/hokenryou.html. a Two different sales figures used to calculate premiums.

below those offered on the private market directly (table 3.4). Apparently, the chambers of commerce negotiate preferential rates for their members. 32 Japanese fi rms pay more claims than those that appear in court statistics, but still not many. The discussion above suggested that Japanese consumers fi le at most 100– 300 products liability suits a year. According to figure 3.4, insurers actually pay 500–1,100 claims a year. Yet the effect on total liability remains trivial. As per table 3.5, a plastic-goods manufacturer can buy insurance on the private market at about 400 yen per million yen in sales. If it sells a gadget for 1,000 yen, the price will include products liability coverage costing 0.4 yen. In US dollars, if it sells a gadget for $12, the price will include insurance costs of 0.5 cents.

Table 3.5 Products liability insurance: Chamber of Commerce rates for small fi rms Coverage Industry Supermarket Textiles manufacture Food retail Household goods retail Household goods manufacture Ceramic product manufacture Construction materials manufacture Glass products Carpentry Plastic manufacture Shoes manufacture Electronic parts manufacture Bakery manufacture Food production Machinery production Restaurant Chemical product manufacture Carpentry tool manufacture Bakery retail Athletic equipment manufacture Auto repair

100 million

300 million

87 95 107 124 133 158 164 173 211 241 268 311 317 338 435 548 549 605 826 1073 1221

129 144 164 187 200 236 228 267 279 357 418 462 493 521 613 854 808 908 1308 1696 1656

Notes: Rates are per one million yen in annual sales. “Coverage” is total personal injury claims paid per year. Source: Okinawa Chamber of Commerce (2011), www.oki-shokoren.or.jp/kyosai/pl/pl.html (2011).

Figure 3.4. Number of claims paid (or about to be paid) under products liability insurance contracts. Contracts are available only to small- and medium- sized fi rms, but are offered to all such fi rms even if they do not sell products covered by the product liability statute. Source: Tokyo kaijo nichido kasai hoken (2010).

58

Chapter Three

Figure 3.4 probably undercounts total claims by about half. Each year, about 1,000 (500–1,100) Japanese assert product liability claims against insured fi rms. But the JCC coordinates this program only for small and medium-sized fi rms. Larger fi rms either negotiate their own policies or self-insure. The Ministry of Economy, Trade and Industry estimates that the smaller fi rms produce about half the Japanese manufacturing output. 33 Consequently, if victims assert 1,000 claims against the insured smaller fi rms, perhaps they assert 2,000 or more (depending on the fraction and type of fi rms that insure) total claims a year.

IV. Why the Low Numbers in Japan? A. Nonreasons 1. introduction Americans and Japanese buy similar products. They use similar services. And they suffer accidents at similar rates. Yet, where Americans claim and sue over accidents often, Japanese rarely do. The question is why: why this difference in the rates at which people demand compensation? Turn fi rst to Japan (section IV) and then to the United States (section V). Within Japan, turn fi rst to factors (routinely raised by observers) that do not explain the claiming rates and then to factors that do. 2. japanese talk first, sue later? Scholars sometimes claim Japanese and Americans switch the order of talking and suing. Americans sue fi rst and talk later. They fi le a suit to show their earnestness and only turn to negotiations thereafter. By contrast, these scholars continue, Japanese negotiate fi rst, and only sue if the talks go nowhere. Americans who plan to pursue a claim approach a manufacturer after they fi rst signal their determination by fi ling suit; Japanese approach a manufacturer at the outset and sue only if the fi rm refuses to deal. Americans fi le suit as a routine step in the negotiations; Japanese sue only after the negotiations fail. Whether true in the abstract, the generalization does not explain any contrast in products liability. At most, it explains only why Japanese plaintiffs fi le fewer suits. It does nothing to explain why they assert fewer claims. Yet, Japanese do not just fi le fewer products liability suits

A System with Few Claims

59

in court. According to the insurance data, they also raise dramatically fewer claims outside of it as well. 3. japanese judges award less? Observers sometimes claim Americans fi le more suits and claims than do Japanese, because American judges award more generous damages. The explanation is wrong on the facts: American judges do not award more generous damages. For the vast majority of disputes, they award amounts that roughly track those in Japan. US judges award substantially less than many scholars claim. For instance, comparativist Mathias Reimann (2003, 807) reported that US judges in 1999 awarded a mean products liability award of $3,045,908. Unfortunately, the figure comes from the commercial Jury Verdict Research Service fi rm (JVRS; Zekoll 2002, 151), and the JVRS misleads. For 1993– 99, the JVRS did indeed report a $3 million mean. Those data, however, included one $285 million verdict. The median for the period was only $608,400. Even the $608,000 median misleads. Deborah Jones Merritt and Kathryn Ann Barry (1989, 326) hand- collected jury verdicts over a twoyear period for Franklin County, Ohio. They then compared their results with the data from commercial jury verdict reporters, including the JVRS. They found: Most troubling of all, the commercial services appeared more likely to report verdicts that favored plaintiffs and, among successful plaintiffs, to include higher recoveries. Plaintiffs won almost one-third (29.3%) of the verdicts reported by the commercial services, but only one-fi fth (19.5%) of verdicts omitted by those services. Similarly, the mean verdict for successful plaintiffs included in the commercial services was $1,483,619. For plaintiffs who prevailed in trials omitted by those services, the mean verdict was less than onesixth that amount: just $236,972.

Economists Alexander Tabarrok and Eric Helland (1999, 171) more plausibly reported 1993 median product liability awards of $260,000. The figure includes more than wrongful death claims, of course. For tort claims in Texas (1988– 2003; “mono-line general liability, commercial auto liability, Texas commercial, medical professional liability, and other professional liability”), Frank Cross and Charles Silver (2006, 1899–

60

Chapter Three

1905) reported a median $199,586 for wrongful death. Their wrongful death awards started at $172,445 for children, rose to $263,864 for victims in their thirties, and fell to $139,693 for those aged over 80 years. 34 By any of these measures—even the high JVRS figure—American courts do not award higher product liability damages in routine cases than do Japanese courts. They apparently grant a few stratospherically high awards—no tort claimant in Japan has ever collected $285 million. But Japanese courts regularly award amounts that are generous and that compare favorably to routine American cases. In products liability suits involving wrongful death, successful Japanese plaintiffs in reported cases collected a mean 61 million yen per case (table 3.2). To reach these figures, judges followed the valuation formula outlined in chapter 2 (contrary to the claims occasionally made, these formulae are not specific to traffic accidents). As table 2.6 details, in those traffic accidents wrongful death awards range from 220,000 to 220 million yen, with a mean 72 million yen and a median 66 million yen. The median 66 million yen tracks almost perfectly the (erroneously high) JVRS median of $608,000. It dramatically exceeds Cross and Silver’s median of $199,586. 4. japanese lawyers cost more? Some observers claim American victims more readily sue and claim because American lawyers are so cheap. In truth, American lawyers are no cheaper than Japanese lawyers. As evidence for the affordability of American attorneys, scholars routinely point to contingency fees. The observation misleads. A contingency fee couples a contract for legal service with a high-risk loan. The contingency-fee attorney does not just sell legal services; he also lends money. The loan is high-risk because the attorney allows the plaintiff to default on the loan should he lose the case. In return for that risky loan, the attorney charges a high interest rate. He charges it implicitly, of course—as the difference between the fraction of the eventual judgment he takes if successful and the market value of the time he invests in the case. Given that the contingency-fee attorney loans money to plaintiffs, he will not take all cases. If a plaintiff’s probability of success seems too low (i.e., if the risk on the loan is too high), he will simply refuse to take the case. Attorneys may offer to represent their clients on a contingency fee basis, in other words, but they do not promise to represent all victims.

A System with Few Claims

61

They agree to represent a victim only if the probability of success is high enough that the expected value of their recovery equals the sum of the market value of their legal services, plus the market interest on the advance of those services. Japanese law does not ban contingency fees, and the bar does not set prices. By custom, litigators tend to couple a fi xed upfront retainer with a contingent percentage of the eventual outcome. The practice is not mandatory, though, and (as the survey results below show) most lawyers charge only a modest initial retainer anyway. Instead, they collect the bulk of their fee through the contingent component. Consider an anonymous bar association survey from 2008 (Nihon bengoshi rengo kai 2009, 12). The survey asked lawyers what they would charge to represent a plaintiff in court over a traffic accident that ultimately yielded a 10 million yen verdict. There were 48.6 percent who replied that they would charge about 300,000 yen as a noncontingent retainer. Another 19.7 percent said they would charge 200,000 yen (with the remainder scattered across other values). The bulk of the fee the lawyers instead took on a contingency basis. Of the respondents, 35.4 percent said that they would charge a contingent fee of 500,000 yen payable at the end of the litigation. There were 18.2 percent who would charge a contingent 700,000 yen, and 15.1 percent would charge 600,000 yen (with the remainder again scattered). Note that the initial noncontingent retainer was half the contingency fee or less, and that the sum of the two totaled less than 20 percent of the recovery. The survey also asked lawyers what they would charge for a more complex medical malpractice claim (Nihon bengoshi 2009, 34). After three years of litigation, posited the survey, the case generated a 10 million yen verdict. Given that the case involved more work, the lawyers did demand a larger noncontingent retainer. Still, the amounts remained modest: 39.8 percent replied that they would charge 500,000 yen, and 26.0 percent said they would charge 300,000 yen (with the remainder again scattered). Again, the lawyers collected the bulk of their fee as a contingent payment at the end: 1 million yen for 46.5 percent of the lawyers, and 1.2 million yen for 20.1 percent. Furthermore, 38.3 percent said they would add another 200,000 yen for the evidentiary work, and 30.9 percent would add 300,000 yen. Note that even in this more complex case, lawyers still charged less than 20 percent of the recovery. For hourly work as well, lawyers with a local community-based practice charge low prices. According to the bar association survey (Ni-

62

Chapter Three

hon bengoshi 2009, 29), 47.0 percent of the community-based lawyers charged 10,000 yen per hour. Another 19.1 percent charged 20,000 yen per hour, 13.6 percent charged 30,000 yen per hour, and 10.6 percent charged 5,000 yen per hour. As befitting these modest fees, Japanese attorneys earn modest incomes. In an anonymous 2010 bar association survey, they reported a mean income of about 14.7 million yen and a median income of 9.6 million yen. Adjusted for age, they just barely earn more than white- collar business executives and substantially less than physicians (Jiyu 2011, 113, 124; Nakazato, Ramseyer, and Rasmusen 2010). In the United States, salaried lawyers (obviously excluding the “AmLaw 500” partners) earned a median salary of $112,760 in 2010 (Bureau of Labor Statistics 2013). Of all 21,000 Japanese attorneys (salaried and partners) in 2004, only 404 paid taxes of over 10 million yen—implying a taxable income of about $390,000. Only one lawyer made at least $6 million (Nakazato, Ramseyer, and Rasmusen 2010). 5. japanese courts deny discovery? From time to time, observers attribute the low level of claiming in Japan to the absence of American-style “discovery” mechanisms. In product liability (and medical malpractice—see chaps. 4 and 5), they argue, victims often lack the evidence they need to prove liability. The would-be defendant controls that evidence but will not disclose it voluntarily. In the United States, a victim could fi le suit and then use the “discovery” process to force the defendant to furnish the evidence. In Japan, because courts do not provide that mechanism, victims cannot force disclosure. They know that the would-be defendants have wronged them, but absent discovery cannot prove the point in court. Victims fi nd compensation foreclosed; malefactors escape responsibility. This story disguises the corrosive effect of American discovery rules. Given the generous approach that most judges take to plaintiff demands, strategic lawyers can (and do) target their discovery demands to those documents that cost the most for the defendant to produce. The logic turns less on what the plaintiff needs to prove his case. It turns more on what the plaintiff can do to impose costs on the defendant. Faced with grinding discovery demands that distract employees from operating the business, even blameless defendants settle. The notion that without discovery wronged victims cannot recover

A System with Few Claims

63

also misses the way judges run trials outside the United States. In a country like Japan (indeed, in most countries throughout the world), a judge controls the trial. He structures it through discontinuous sessions that he holds about once a month. At each session, he hears evidence, discusses the case, and tells the parties what evidence to bring to the next session. If a plaintiff can convince him that the defendant controls access to evidence the plaintiff needs, the judge can simply tell the defendant to bring the evidence to the next meeting. Plaintiffs cannot conduct “fishing expeditions,” to be sure, but if they need information the judge can tell the defendant to provide it. To induce defendants to comply with these orders, judges use burdens of proof and presumptions (Huang 2003, chap. 3; Uchida 2007, 328). If a judge thinks that the defendant holds crucial evidence, he can switch the burden of proof. The judge can “presume” a fact against the defendant unless he introduces enough evidence to rebut the presumption. Recall the exploding Panasonic television set (sec. II.A, above). 35 TV sets should not explode, the judge observed. This one had. The plaintiff could not prove that Panasonic had been negligent, but the judge did not care. Given the explosion, he would presume that Panasonic had been negligent. To avoid liability, Panasonic could prove the contrary. When Panasonic failed to do so, the judge declared it liable. Judges adopt much the same strategy in medical malpractice cases (see chaps. 4 and 5). 6. japanese courts deny aggregation? Some observers sometimes attribute the high US litigation rates to the aggregation rules: in particular, to class actions. Suppose no single victim loses enough to warrant a suit, but many people suffer similar types and levels of injury. Under the class-action rules, they can recover together. One of them will fi le suit on behalf of them all. If he wins, they all share in the judgment. Japan does not have class actions, and neither do most other countries (Reimann 2003, 819– 20). Instead, judgments bind victims only if they appear as named plaintiffs. The putative theory behind the class-action mechanism was simple: fi rms sometimes cause massive damage to society at large, but only trivial injuries to any one victim. Litigation entails fi xed costs. If each victim suffers only small damages, no one will have enough at stake to justify paying those fi xed costs. If no one sues, the defendant will escape liability even when he causes huge damages overall. Let attorneys sue on

64

Chapter Three

behalf of all victims, and they can overcome the fi xed costs that would other wise prevent recovery. Unfortunately, the logic misleads badly. To be sure, because each plaintiff has only a small stake in the action, no one has an incentive to sue. But because each has such a small stake, no one has an incentive to monitor the lawyer either. Instead, the class-action attorney operates as an independent, unmonitored agent. He then negotiates a settlement with the defendant that is generous to himself and spartan to his clients. He and the defendant propose the settlement to the judge, and the busy judge approves it. The most egregious of the class actions may be the securities class action. The number of cases is small: from 1996 to 2009, the number of securities class action suits exceeded 300 only once. Yet, to settle these cases fi rms pay dearly. In 2009, plaintiffs fi led 221 securities class actions and settled cases for a mean $13 million, $2.8 billion overall. Of that amount, the total to the attorneys: $963 million (Milev et al. 2010, figs. 1, 18, 19, 20). The defendants settle these cases because the class-action rules ramp the amounts at stake to such high levels that they can litigate them only by betting the firm. Firms that do litigate them rarely lose. From 1995 to 2010, plaintiffs filed 3,400 securities class actions in federal court. Only 27 of the 3,400 went to trial, and of those that did the plaintiffs won 6 and obtained a mixed verdict in 5 (Milev et al. 2010, 16, 19, fig. 13). As Judge Richard Posner put it: “because of the astronomical damages potential of many class action suits, a grant of certification may place enormous pressure on the defendant to settle even if the suit has little merit.”36 Ultimately, he explained: “certification as a class action can coerce a defendant into settling on highly disadvantageous terms regardless of the merits of the suit.37 Although the US class-action rules do encourage meritless (indeed, fraudulent) litigation, they do not explain the US-Japan contrast in products liability: American lawyers simply bring too few class actions to explain that difference. Remember that there are 80,000 products liability suits in the United States, but barely 100– 300 in Japan. Of the 80,000 suits, only a trivial number are class actions. Brian Fitzpatrick (2010, 818) located 304 federal class action settlements in 2006 and 384 in 2007. Of these, he listed 40 of the cases in 2006 and 47 of the 2007 cases as “consumer” class actions (and not even all these were product liability cases). In addition to the federal cases, he noted that plaintiffs typically fi le a larger number of class actions in state court. With forty federal

A System with Few Claims

65

consumer class actions, even if plaintiffs fi led ten times more class actions in state court, the number would not begin to explain the difference between the United States and Japan. B. An Alternative Explanation When observers propose these hypotheses about the low claiming and litigation numbers in Japan, they miss the more obvious explanation: Japanese bring few suits because they suffer few accidents. The point is very simple: Japanese seldom claim or sue over defective products because they seldom suffer damages from defective products. Given the number of serious product-related accidents, Japanese fi le about as many claims involving those products as one would expect. Take figure 3.1 again: in 2010, 33,510 Japanese died of nontraffic-related accidents, but the vast majority of these were aged >65 years. Elderly people who slip in the bathroom or fall down the stairs do not slip or fall because of any product defect. They slip or fall because they lost their sense of balance. In the same year, only 6,651 younger Japanese died of nontraffic-related accidents. Yet, just as most of the old people who died of accidents did not die from product defects, neither did most of the 6,600 younger victims. Every year 800–1,200 Japanese swimmers drown, usually in the ocean. 38 Another 200–400 people die mountain climbing. 39 Infants slip off second-floor balconies. Grade-schoolers fall through ice. Middleschoolers crash their skateboards down a hill. Vacationing twentysomethings ski into trees. Every year, 6,600 nonelderly Japanese die in nontraffic-related accidents, and another several thousand suffer serious injuries but do not die. Only some of these victims die or suffer from defective products. Given these numbers, if Japanese assert 2,000 product- defect claims a year (section III.C, above), they probably assert most of the substantial product liability claims they have.

V. Why the High Numbers in the United States? A. Introduction The comparative puzzle concerns not the low level of claiming in Japan but the high level in the United States. Japanese assert about as many

66

Chapter Three

claims as the number of serious product-related accidents. Americans buy the same products and suffer roughly the same number of accidents. The puzzle is why they fi le so many more claims and suits. Take some simple numbers. In a typical year, Americans fi le 80,000 products liability claims (see section III.B, above). Yet in that same typical year, only 83,000 Americans die from any nontraffic accident.40 Most of them are old—and a 75-year- old woman who falls and breaks her hip does not fall because of any product defect. In a typical year, only 49,000 Americans aged under 65 years die from nontraffic-related accidents.41 As with young Japanese, even most of these Americans do not die from defective products. After all, 33,000 of the 49,000 died from poisonings—rarely poisoning from a defective product. Another 3,800 Americans drown—again, rarely from a defective pool or bathtub, much less a defective lake. And as in Japan, people suffer serious injuries or die from bad luck and bad judgment: students die playing football; skaters fall through the ice; skiers run into trees; and babies fall off balconies. In a typical year, far fewer than 49,000 Americans die from anything defective. Yet 80,000 Americans fi le products liability suits anyway. At root, Americans fi le so many products liability claims because American procedure so generously rewards meritless claims. Litigation rates are not lower in Japan because of a Japanese pathology: Japanese law does not block valid claims. Instead, rates are higher in the United States because of a distinctly American dysfunction: American law rewards claimants who pursue compensation to which they are not legally entitled. Rates are higher in the United States because plaintiffs (whether or not actual victims) can manipulate the courts to extort payoffs from defendants (who may or may not have done anything wrong). B. Jury Trials Where American plaintiffs try product liability (and other tort) cases before juries, Japanese plaintiffs try them before professional judges. The contrast reflects the difference between fi rst- and second-best approaches: the fi rst best assigns fact fi nding to the “jury of one’s peers”; the second best hires a professional judge. In using judges, Japan is no outlier. Civil law countries have long avoided civil juries, and today even most common-law countries do. New Zealand “has relegated [them] to only one or two cases per year” (Cam-

A System with Few Claims

67

eron et al. 2000). Canada retains civil jury trials “in some jurisdictions in name only,” and in other jurisdictions has abolished them entirely (Bogart 2000). Even in the United Kingdom, “less than one percent of civil trials” use juries, and in personal injury cases plaintiffs have no right to a jury at all (Lloyd-Bostock and Thomas 2000). The civil jury is central to the problems that plague American products liability. “Juries are a great attraction for plaintiffs,” noted Reimann (2003, 813–14): In the end, the most important concern about juries is not what they do but what they might do. Their unpredictability threatens a losing defendant with a potentially catastrophic verdict. Rather than gamble, most manufacturers or sellers prefer to avoid the risk and settle, even at high cost.

Aggressive attorneys do not fi le products liability cases randomly. Instead, they pick the courts where they know juries award the most inflated recoveries. The Scruggs of cigarette and asbestos fame called them “magic” jurisdictions (Boyer 2008): The trial lawyers have established relationships with the judges that are elected; they’re State Court judges; they’re populists. They’ve got large populations of voters [who serve on the juries] who are in on the deal, they’re getting their piece in many cases. And so, it’s a political force in their jurisdiction, and it’s almost impossible to get a fair trial if you’re a defendant in some of these places. . . . The cases are not won in the courtroom. They’re won on the back roads long before the case goes to trial. Any lawyer fresh out of law school can walk in there and win the case, so it doesn’t matter what the evidence or law is.

These “magic” juries (who typically come from very poor communities) do not just award large amounts. Instead, they award the largest amounts to local plaintiffs against out- of-town defendants (Helland and Tabarrok 2003, 38, and 2006, chaps. 2 and 3). Helland and Tabarrok (2003; 2006) made the point most starkly. They noted that defendants in traffic cases are mostly local, while products liability defendants tend to be fi rms from other states. They then compared verdicts that juries awarded in wealthy and poor communities. Plaintiffs from richer counties (poverty rates of under 5 percent) receive a mean $244,000 award in automobile accident disputes and $1.18 million in products liability dis-

Chapter Three

68

putes. Plaintiffs from poorer counties (poverty rates over 25 percent) receive $760,000 (3.1 times as much as in the rich counties) in automobile accident disputes but $6.74 million (5.7 times) in products liability disputes Scruggs’s “magic” jurisdictions were poor jurisdictions. Their juries do not just dispense high awards; they dispense astonishingly high awards against the nonlocal defendants. C. Elected Judges Where Japanese run their trials before judges appointed by the national government (effectively, bureaucrats—see chap. 7), Americans run products liability trials before popularly elected state judges (again, a fi rstbest philosophy tied to “democratic accountability”). Who decides a judge’s tenure matters. Legislators answer to local voters —and so do elected judges. If jurors want to redistribute assets from out- of-state corporations to their neighbors, some state judges declare themselves happy to oblige. As one retired West Virginia Supreme Court justice observed (Neely 1988, 4): The anarchy that currently prevails among American state jurisdictions absolutely guarantees politically that no line of any sort will be drawn. After all, I’m not the only appellate judge who wants to sleep at night. As long as I am allowed to redistribute wealth from out- of-state companies to injured instate plaintiffs, I shall continue to do so. Not only is my sleep enhanced when I give someone else’s money away, but so is my job security, because the instate plaintiffs, their families, and their friends will reelect me.

Local juries vote, and out- of-state corporations do not. The same judge continues, in a discussion of a particular case (Neely 1988, 70– 72): From what I know about myself and my colleagues, I have the distinct impression that in a product liability case the vote would have been 3 to 2 the other way, and the whole $10 million judgment would have been sustained. Had a defective Ford automobile killed the little boy, even I would have had none of the enthusiasm for reducing the judgment that I had when the judgment against the defendants would affect business and consumer costs in West Virginia. What do I care about the Ford Motor Company? To my knowledge Ford employs no one in West Virginia in its manufacturing processes. . . . The

A System with Few Claims

69

best that I can do, and I do it all the time, is make sure that my own state’s residents get more money out of Michigan than Michigan residents get out of us.

Kip Viscusi (1991, 56) quoted another state judge along much the same lines: I may not always congratulate myself at the end of the day on the brilliance of my legal reasoning, but when I do such things as allow a paraplegic to collect a few hundred thousand dollars from the Michelin Tire company—thanks to a one- car crash of unexplainable cause—I at least sleep well at night. Michelin will somehow survive (and if they don’t, only the French will care), but my disabled constituent won’t make it the rest of her life without Michelin’s money.

Elections affect outcomes. Compared to the outcome in a state court that selects judges through a nonpartisan mechanism to the outcome in a court that selects judges through partisan elections, the probable award is 23 percent higher (Helland and Tabarrok 2006, 88– 90). Eric Helland and Alexander Tabarrok (2006, 92) explain: In cases involving out- of-state defendants and in-state plaintiffs, the average award (conditional upon winning) is $362,988 higher in partisan states than in nonpartisan states; $230,092 of the larger award is due to a bias against out- of-state defendants, and the remainder is due to generally higher awards against businesses in partisan states.

The American civil procedural system sometimes lets attorneys use civil juries to extract large payments from out- of-state defendants; elected judges sometimes acquiesce to the scheme; and the idiosyncratic spectacular verdict and class action magnify the combined effect of those two factors.

VI. Conclusion Compared to traffic accidents, products liability suggests a darker, more abusive logic to American litigation patterns. Litigation rates are not higher in the United States because American courts more effectively

70

Chapter Three

help victims of defective products recover their losses. Instead, they are higher because judges facilitate extortion by the plaintiffs’ bar. American judges do not help victims recover losses more effectively than do Japanese judges; American attorneys are not more affordable; median American judgments are not higher; discovery does not give victims information they would lack in Japan; and class actions comprise only a trivial number of suits. Where the Japanese legal system combines simple principles consistent with its second-best approach to judging, the US system is complex in a way that facilitates abuse. It is complex precisely because of its fi rstbest insistence on fitting particularized adjudication to the specifics of each case. Real-world courts with fi nite budgets and human judges cannot reliably tailor adjudication to the individual idiosyncrasies. American courts purport to try—and the result is what we see: small numbers of spectacularly high judgments that badly skew the mean; punitive damages that potentially raise idiosyncratic awards even higher; class actions that let strategic attorneys turn trivial injuries into massive claims; and discovery that lets a plaintiff’s attorney impose relentlessly grinding demands. Together, these features of the US legal system let abusive attorneys fi le claims so large that risk-averse managers settle even meritless claims rather than bet the fi rm. Perhaps the defendant did no wrong; perhaps the victim suffered no harm; or perhaps the defendant simply caused far less harm than the plaintiff claimed. The plaintiff knows he fi led an abusive claim; the defendant also knows; and the court ought to know, too. But in American litigation nothing is certain. Rather than hope that the court will sort the evidence as it should and reach the right result, defendant managers all-too- often just settle in order to make the problem go away.

Chapter Four

Few Claims, but for a Different Reason Medical Malpractice (I)

T

he evidence in chapter 2 suggests that victims of serious traffic accidents (or their heirs) almost always assert their claims. They assert them privately, but in the shadow of the law. Out of court, they settle by reference to what a judge would award if they sued. They collect generous amounts, but can negotiate the payments privately because Japanese judges give clear signals about what they would do if they did litigate the claims. The material describes a second-best system that works. The Japanese court system is bureaucratized (see chap. 7)—but also clear and generous. It is clear because it is bureaucratic; its clarity renders it predictable; and that predictability lets victims negotiate their settlements privately. In the process, everyone saves the money he would otherwise pay his lawyer. Chapter 3 describes the world of products liability, and at least superficially it describes a very different world. Japanese fi le many fewer products liability suits than Americans, and they assert fewer claims out of court too. Observers typically attribute the scarcity to a dysfunctional legal system: trials take too long; lawyers charge too much; judges award too little; courts offer no discovery; class actions are unavailable. Yet none of these explanations works: none explains why Japanese pursue so few products liability claims, and most are not even true in the abstract. Although Japanese pursue fewer products liability claims than Americans pursue, they do not pursue fewer claims than they have. Instead—

72

Chapter Four

exactly as in chapter 2—victims with serious claims assert them. Generally they settle, but sometimes they litigate. When they do, they fi nd that the average trial does not take longer than in the United States. They fi nd that lawyers are cheaper than in the United States; that the average judgment is more generous; and that judges use means other than discovery to give them the information they need. True, they cannot file class actions, but class actions do not explain the large number of American suits anyway. Americans fi le so many more claims only because the US legal system takes a fi rst-best particularized approach that fails to account for the ordinary human limitations of judges and juries—and ultimately rewards nonvictims with fraudulent claims. Japanese also fi le many fewer medical malpractice suits than Americans fi le and assert many fewer claims out of court. Again, observers attribute the small number to a dysfunctional Japanese court system. Yet that the system successfully handles both traffic accidents and defective products suggests we should hesitate before concluding that systemic defects cause fewer claims. Perhaps, as in products liability, American claiming rates are high only because nonvictims fi le fraudulent suits. Perhaps—but that is not what health- care specialists say. Scholars of the US health care system do indeed fi nd some plaintiffs who sue nonnegligent doctors. But they also describe victims who let clearly negligent doctors go unsued. The field includes both type I and type II errors, they conclude: some patients assert negligence where there is none, and others ignore negligence where it exists. In chapters 4 and 5, I argue instead that Japanese plaintiffs fi le fewer claims because they see less malpractice. They do not see less malpractice because the quality of medical care is high. Perversely in the extreme, they see less of it because the quality is low. Physicians most often commit malpractice when undertaking difficult, high-risk, complex procedures on the frontiers of science. They seldom commit it while prescribing ordinary antibiotics during a routine visit or setting a broken bone after a bicycle accident. Japanese doctors do much less of the former than do American doctors. They do more of the latter. The nature of Japanese health care reflects the nature of the funding: the Japanese government operates a national health insurance system, and operates it on the cheap. To keep it cheap, the government reimburses doctors at low rates. In the process, it generates the quality degradation that price controls inevitably cause. Yet, crucial to the

Few Claims, but for a Different Reason

73

question of malpractice, it suppresses costs differentially: although it reimburses physicians for all procedures at low prices, it suppresses the prices for the difficult, high-risk, complex procedures more severely than the prices for the ordinary antibiotics and bone fractures. Predictably, given this policy, Japanese receive very little care at the scientific frontier and vast quantities of routine treatment. Relevant here, they receive little of the type of care that generates the most malpractice claims and large amounts of treatment that seldom does. In chapters 2 and 3, the contrast in US-Japan disputing patterns reflected the difference between fi rst- and second-best approaches to adjudication. The contrast in medical malpractice disputing patterns does not reflect that difference. Indeed, it does not reflect a difference in the court system at all. Instead, it reflects different patterns of health- care delivery driven by different mechanisms of funding. I begin by reviewing some basic comparative statistics on medical malpractice, as well as the secondary literature on malpractice litigation in Japan (section I). I offer a brief description of the Japanese health care industry (section II). I explore two subsidiary consequences of the way that the Japanese government has kept prices low (sections III and IV). I then return to the malpractice litigation in chapter 5.

I. The Literature A. United States: The Short Story In the United States, patients or their heirs fi le 50,000–160,000 paid malpractice claims each year. Lawyer, physician, and public-health scholar Michelle Mello and coauthor David M. Studdert (2006, 13) proposed the low end of the range: 50,000– 60,000 paid claims annually, for a total transfer of $5.8 billion. The AON insurance brokerage fi rm proposed the high end: 156,000 paid claims, for a total $28.7 billion (AON 2004). Within the United States, disputing patterns vary widely by region. From 1990 to 1997, claimants in Florida (with its fourteen million population) fi led about 2,600 claims and collected on 50– 60 percent of them. Those in Texas (population twenty-four million) in 2002 fi led 6,929 claims and collected on 5,555. At trial, US malpractice claimants prevail perhaps 20– 30 percent of the time.1 As Mello and Studdert (2006, 13) put it:

Chapter Four

74

A reasonable estimate [of the number of malpractice claims brought each year] is probably in the 50,000 to 60,000 range. Available figures suggest that approximately 70 percent of malpractice claims do not reach trial. Those that do are heard by a jury. . . . Plaintiffs prevail in approximately 30 percent of trials. Considering settlements and verdicts together, about 30 percent of all claims are closed with a payment to the plaintiff.

Successful US patients (or heirs) collect $150,000–$310,000 each. AON suggested that the average successful claimant collects $178,000. In Florida, the median successful claimant in 2003 collected $150,000 and the mean claimant $300,000. An insurer trade association reported median 2001 payouts of about $180,000 and mean payouts of $310,000. And Mello and Studdert estimated the range for 2003 at $260,000–$310,000. 2 In wrongful death claims (25– 35 percent of the claims), heirs collect $200,000–$300,000. In Florida, they received median payments of $195,000 and mean payments of $290,000 (1990– 2003). According to the National Practitioner Data Bank (NPDB; operated by the Department of Health and Human Services), in 2005 they collected median payments of $175,000. 3 To protect themselves against malpractice claims, physicians buy insurance. Insurers price the premiums according to the insurance underwriting cycle, location, and specialty, but not to the history of claims against a given physician. As of 2000, the mean American doctor paid a premium of about $18,500 (Bovbjerg and Bartow 2003, 13). In some states doctors pay more: the mean physician in West Virginia paid $39,050. And in some specialties they pay more: in orthopedic surgery, the mean American doctor paid $38,200 (ibid., 15). In fact, premiums can cripple. Obstetricians and surgeons have struggled particularly hard, and regional shortages have ensued. In 2009, surgeons in Massachusetts faced premiums of $42,000 per year. Orthopedic surgeons in Pennsylvania paid $73,000. Obstetricians in Massachusetts paid $101,000. Surgeons in Dade and Miami County, Florida, paid $190,000 per year, and obstetricians paid $202,000.4 No one claims the suits hit the doctors they should. Some plaintiffs sue doctors who did no wrong, while many victims of malpractice apparently never fi le a claim. 5 Despite the high costs involved, American insurers steadfastly refuse to experience-rate their physician clients: seemingly, past liability does not predict future claims. In medicine as in other fields, some people are smarter than others, and some are more

Few Claims, but for a Different Reason

75

careful than others. If litigation identified genuine malpractice, insurers should fi nd past liability a useful predictor of future exposure. Given the bizarre mix of type I and type II errors, they seem not to do so. Nonstandardized and complex, malpractice claims take time to resolve. From the time of the accident to the eventual resolution of the dispute (litigated and settled claims grouped together), claimants in Florida spent about 3.3 years (Vidmar et al. 2005, 330– 31, tables 1 and 2). In the United States more broadly, they spent a median 4.13 years and a mean 4.66 years (NPDB 2005). B. Japan: The Data The gist of medical malpractice litigation in Japan is clear: plaintiffs fi le few claims in court; of the claims they do fi le, they drop or settle most; if they litigate, they spend about three years in court; and at their fi nal disposition, they win about 30–40 percent of the cases. In 2004, Japanese plaintiffs fi led about 1,100 medical malpractice cases. In the same year, the district courts closed about 1,000 (table 4.1, panel A). Although these numbers are low, they did not represent a decline. Instead, as table 4.1 shows, they were nearly twice as large as they had been in 1998. After peaking in 2004, the number of suits fell a bit—to 707 in 2009. In 2012, plaintiffs fi led 770 malpractice suits (Saiko saibansho, Iryo, 2013, 2). The parties litigated about 40 percent of these suits to a fi nal (lowercourt) judgment. Plaintiffs fi led 632 suits in 1998, and the courts adjudicated 232 (table 4.1, panel A). They fi led 1,110 in 2004, and the courts adjudicated 405 (40 percent). In the rest of the cases, either the plaintiffs dropped their claims or the defendants paid out of court. These settlement rates track those for civil litigation more generally: the courts closed 149,000 ordinary litigation cases in 2004, but adjudicated only 71,000 (48 percent) (Saiko saibansho, Shiho tokei 2004, table 20). They closed 168,000 ordinary litigation cases in 2012, but adjudicated 69,750 (41 percent) (Saiko saibansho, Shiho tokei 2012, table 20). Table 4.2 provides information about the types of plaintiffs and defendants involved in these malpractice cases. Malpractice claims take longer to try than the typical civil suit. In 1998, from fi ling to judgment the courts in malpractice suits took thirtyfive months.6 By 2004, they had cut that number to twenty-seven months (table 4.1, panel B), and by 2006 to twenty-six months (Saiko saibansho,

Table 4.1 Selected summary statistics of medical malpractice cases (I) A. Numbers of decisions All cases 1 Cases fi led 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

2 Cases closed

632 678 795 824 906 1,003 1,110

3 Court decisions

582 569 691 722 869 1,035 1,004

232 230 305 334 386 406 405

4 Published opinions 42 44 33 38 23 48 44 30 30 16

Notes: Columns 1–3 are from the administrative office of the courts. Column 4 is from the published opinion database, as described in the text. Column 3 are hanketsu judgments.

B. Filing-to-judgment times and recovery rates 1

2

3

All cases Filing to judgment, years 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

2.93 2.88 2.97 2.72 2.58 2.31 2.28

4

Published opinions

Plaintiff recovers, %

Filing to judgment, years

Plaintiff recovers, %

43.5 30.4 46.9 38.3 38.6 44.3 39.5

5.00 3.80 4.37 4.86 5.00 4.24 3.68 3.55 3.82 3.55

66.7 65.9 60.6 71.1 73.9 72.9 75.0 86.7 96.7 93.8

Notes: Column 2 includes only cases proceeding to judgment (hanketsu); column 1 includes cases that settle. Columns 1 and 2 are from the administrative office of the courts; and columns 3 and 4 are from the published opinion database, as described in the text.

C. Principal litigation venues, % 1 Published medical malpractice Tokyo Osaka Nagoya Yokohama Fukuoka Kobe Other

30.5 16.4 4.9 4.9 4.3 3.4 35.6

2 Ordinary civil litigation 20.5 9.4 4.3 3.5 9.5 3.0 49.7

3 All medical malpractice 40.6 26.5 5.5 — 8.1 4.2 20.1

Sources: Published opinion database, as discussed in the text; Maeda, Sakamoto, and Nobutomo (2001); Inoue (2007); Saiko saibansho, Shiho tokei (2004), www.courts.go.jp. Notes: Data are percentages. Column 1 is from the published opinion database; columns 2 and 3 are from the administrative office of the courts.

Table 4.2 Selected summary statistics of medical malpractice cases (II) A. Principal malpractice claims, by procedure, % Published opinions

All cases

46.3 15.5 16.7 15.2 12.4 10.3 9.8 7.5

38.9 13.7

Surgery related Obstetrics Cancer Medication related Misdiagnosis Cerebrovascular Emergency facilities Cardiovascular

Note: Published opinion dataset, as described in text (n = 348). Categories are not mutually exclusive.

B. Malpractice claims and beds, by institution, %

Government hospital University hospital Other public hospital Private hospital Private clinic Dental clinic

Claims

Beds

20.7 17.5 17.1 22.7 18.7 1.5

21.4 5.7 17.8 55.2a

Note: Claims data from published opinion dataset, as described in text (n = 343). a Includes both private hospitals and private clinics.

C. Malpractice claims, by patient age Age, years 0–2 3–10 11–20 21–30 31–40 41–50 51–60 61–70 ≥71

Civil cases, % 16.1 4.0 6.6 12.6 7.8 10.6 16.1 10.9 15.2

Sources: Published opinion database, as described in the text; www.courts.go.jp/ saikosai/about/iinkai/izikankei/toukei_01.html (through _04.html); Kosei rodo sho (2005). Note: Published opinion dataset, as described in text (n = 348).

78

Chapter Four

Iryo, 2013). Most civil suits take far less time: of the 71,000 suits adjudicated in 2004, the courts closed 76 percent within one year. They closed 97 percent within three (Saiko saibansho, Shiho tokei 2004, table 20). Of the 70,000 adjudicated in 2012, they closed 81 percent within one year and 99 percent within three (Saiko saibansho, Shiho tokei 2012, table 20). Plaintiffs recover damages in about 30–45 percent of the medical malpractice cases they litigate to a fi nal judgment.7 This is lower than the comparable figure for civil litigation more generally. Of the 71,000 ordinary civil suits adjudicated in 2004, plaintiffs won (in whole or in part) 84.1 percent. Of the 44,000 suits in which they demanded a money judgment, they recovered some amount in 80.9 percent (Saiko saibansho, Shiho tokei 2004, table 19). Of the 70,000 ordinary civil suits adjudicated in 2012, they won 84.4 percent, and of the 45,000 suits where they demanded a money judgment they recovered some amount in 81.7 percent (Saiko saibansho, Shiho tokei 2012, table 19). C. Japan—The Scholarship As with products liability, observers often explain why Japanese patients claim as they do by positing a legal system that does not work. Although law professor Robert B. Leflar (2009b, 444–45) noted that the predictability of the damage awards facilitates out- of- court settlement (see also Leflar and Iwata 2005), he argued that the small number of lawyers (about 24,000) raises the cost of malpractice litigation. The “delay in case resolution (at least before recent reforms)” reduces its return. And the tendency of Japanese attorneys to demand a noncontingent retainer at the outset requires many would-be plaintiffs to front cash they do not have. Legal scholar Eric Feldman (2009, 257– 58) similarly noted the predictability of damage amounts, but attributed the dearth of suits to the structural “barriers” that “inhibit access to the legal system.”8 Oddly (given the number of tort claims), he (2009, 259) argued that the “government’s long-standing approach to tort-related claims” had “effectively shut the door to tort litigation.” He cited the noncontingent retainer arrangements (ibid., 264), and noted the time required to litigate the claims (ibid., 269): “in 2006 it still took an average of 25.1 months for the average malpractice case to move from fi ling to fi nal judgment in the district courts.” Where Leflar and Iwata (2005, 200) observed that “mean and median awards in U.S. wrongful death cases . . . seem

Few Claims, but for a Different Reason

79

not to diverge radically from the Japanese scale of things,” Feldman (2009, 265– 66) found Japanese awards low: “more modest than in the United States.” And oddest of all, he faulted Japanese courts for requiring plaintiffs to “prove the central elements of their allegations” (2009, 263– 64). “[B]y requiring plaintiffs to bear the burden of proof in medical malpractice cases,” he explained, “Japanese courts effectively limit the number of malpractice claims that can succeed.” Scholars have also focused several other aspects of the Japanese malpractice disputing environment. For example, in order to gauge the level of claiming, Leflar (2009a, 2009b) and Leflar and Iwata (2005, 201) suggested looking to premium levels. Malpractice insurance premiums, they noted, “could be considered a very rough-hewn proxy for liability payouts in the long term.” They then observed (2009a, 8n28) that physician members of the Japan Medical Association (JMA) obtain their coverage for 70,000 yen (about $700) and general hospitals for about 30,000 yen (about $300) per bed.9 Leflar (2009a) and Leflar and Iwata (2005) also noted that Japanese prosecutors bring criminal charges against physicians for the most egregious cases of malpractice. They rightly noted that the prosecutors bring relatively few claims, but observed that they obtain broad news coverage for the few they do fi le. They then asked whether criminal sanctions might fi ll a gap in incentives left by the small numbers of private suits. Two sets of scholars examine unpublished as well as published decisions in malpractice cases. First, Maeda, Sakamoto, and Nobutomo (2001) studied 310 malpractice cases from 1989 to 1998 in the major district courts (importantly, including Tokyo and Osaka). Like Leflar, Iwata, and Feldman, they attributed the low litigation rate to the delays and attorney fee structure. Second, Hagihara, Nishi, and Nobutomo (2003) examined 435 cases from 1986 to 1998 in the major district courts (apparently a database that overlaps with that of Maeda, Sakamoto, and Nobutomo 2001). They found that 47 percent of the cases involved wrongful- death claims, that 32 percent resulted in a plaintiff recovery, and that the successful claimants in the wrongful- death cases collected a mean 28 million yen. D. Malpractice Claims in Other Countries Perhaps because malpractice in most societies outside the United States involves small aggregate transfers, we know less about the claiming pro-

80

Chapter Four

cesses in other countries. In most advanced democracies, patients fi le far fewer claims than in the United States. But—as in Japan—in most of these countries the government has also for years more closely controlled the medical services industry. The United Kingdom and Canada couple a universal health care program with legal systems otherwise similar to that in the United States. Both have little malpractice litigation. In the United Kingdom, the National Health Service estimates its annual expenditures for malpractice at $642 million (fiscal 2001– 2, on a population of 59 million; Wheat 2005). In Canada, on a population of 32 million, plaintiffs fi led 1,083 malpractice suits (in 2004; CHSRF 2006). Trebilcock, Dewees, and Duff (1990, 542) estimated that “the average frequency of claims fi led against physicians in the U.S. is about five times greater than in Canada.” In several other advanced democracies, the government has displaced the tort regime from malpractice entirely. In New Zealand, Sweden, Denmark and Finland, for instance, it has imposed no-fault instead (OECD 2006, 13–14).

II. Japanese Health Care Industry A. Universal Health Insurance The Japanese government adopted a national health insurance scheme in the early 1960s. The economy was growing rapidly, but so was the electoral threat posed by the socialist and communist left. In their bid for power, these parties championed European welfare state policies. Strategically to trim their appeal, the center-right government fashioned from existing government health plans a universal health insurance scheme of its own. In the half- century since its adoption, the scheme has evolved, but several sets of insurance plans lie at its core. One set of policies, covering 30.6 million people in 2004 (Kameoka 2005, 8–13), comprises a network of employer-based plans at large fi rms. Within this network (mandatory for all fi rms that fall within its scope), employers pay insurance premiums on behalf of their employees. Through the insurance, those employees and their dependents then obtain most health care. A second program covers employees at smaller fi rms (35.9 million). Still other plans cover the self- employed, the unemployed, the retired, and so forth. The insurers for these plans range from private fi rms (for the larger

Few Claims, but for a Different Reason

81

employers) to the Japanese government (for smaller employers) to municipal governments (for the self- employed and the unemployed). Copayments run from 10 to 30 percent, depending on the plan, but were capped in the mid-2000s at 140,000 yen plus 1 percent of the excess over 466,000 yen (less for low-income patients). Because the government uses revenue from the employer-based plans to pay for the others, crosssubsidies are large (Kameoka 2005, 8–13, 42). Through this national health insurance, Japanese obtain a wide panoply of services. Of the 31 trillion yen in 2001 health- care costs, outpatient charges made up 41 percent of that total, and hospitalization costs added another 37 percent (Nakamura 2006, 32). Personnel costs (e.g., doctors and nurses) constituted 49 percent of the health care costs, and pharmaceuticals added another 20 percent. Patients aged under age 65 generated 46 percent of the health- care costs, and those over 75 consumed 35 percent (Nakamura 2006, 19, 32). Every other year, the Ministry of Health, Labour, and Welfare (MHLW) publishes an elaborate price schedule. It minutely details the prices providers may charge for medical, hospital, and pharmaceutical treatments. In setting the prices, its bureaucrats primarily negotiate with the JMA (Campbell and Ikegami 1998, chap. 6). These prices are mandatory. For services billed under the national health insurance, providers may not charge more. Other than a few exceptions for services like private hospital rooms, neither may they charge patients supplementary amounts. And the set prices are low. According to John Campbell and Naoki Ikegami (1998, 147), the prices average about a quarter of the reimbursement rates in the United States. Apparently, this system does keep costs down. As of 2003, the Organisation for Economic Co- operation and Development (OECD) estimated that Japan spent 8.0 percent of its gross domestic product (GDP) on insured health care. Canada spent 9.9 percent, France 10.4 percent, Germany 10.9 percent, and the United States 15.2 percent.10 As of 2010, Canada spent 11.4 percent of GDP on health care, France 11.7 percent, Germany 11.5 percent, Japan 9.6 percent, and the United States 17.7 percent.11 Given the inevitable variation in medical service quality, basic logic suggests that high- quality physicians and high-reservation– price patients should selectively supplement the set prices with side payments. They may. According to one estimate, patients pay about 389 billion yen in bribes, or about 1.3 percent of the total health expenditures.12 Because

82

Chapter Four

these payments go to the few most talented physicians, they may constitute a sizeable fraction of income for those who receive them. For the patients who pay them, they obviously constitute a considerable burden as well. A few patients and doctors do abandon the system entirely. The law does not stop physicians from charging higher prices if patients will forgo the insurance and pay the full amount out of pocket. Perhaps increasingly, they do. Unfortunately, clinics that do not bill the government do not appear in the government statistics. A few clinics catering to expatriates (along with many higher- quality dentists) have operated outside the system for years. Some physicians with a domestic clientele do so as well. B. Physicians Over 259,000 licensed physicians practiced in Japan in 2004 (284,000 in 2010).13 At 2.0 doctors for every 1,000 potential patients (2.2 in 2010), this 2004 figures falls below those for comparably wealthy countries: 2.1 for Canada, 2.3 for the United Kingdom, 2.4 for the United States, 3.4 for France, and 3.4 for Germany (Nihon iryo seisaku kiko 2007). Physicians, in short, are few. Incumbent doctors complain of excess supply, but physicians are scarce because the government limits entry. As in the United States, would-be physicians must fi rst enter a university medical department. Once admitted, they study six years and take a national licensing examination. As approximately 88 percent of the graduates pass, the bottleneck is the medical school requirement rather than the licensing exam.14 Before opening a practice, new doctors work at least two years in a hospital-based residency program.15 Physicians fall broadly into two groups: (a) the men and women who run the small, often low- quality clinics, and (b) the doctors who staff the larger hospitals and sometimes offer very high- quality care. Japanese physicians earn modestly high incomes. According to the MHLW,16 in 2005 salaried physicians earned a mean of 10 million yen. In the United States (according to the US Bureau of Labor Statistics),17 a new pediatrician in 2004 could expect a median income of $133,000; an anesthesiologist with over a year’s experience could expect $322,000. Among Japanese physicians, those who make the most run one of the small private clinics. In effect, they sell expensive hotel stays (typically to elderly patients) at government expense. Predictably, given the

Few Claims, but for a Different Reason

83

incentives involved, they keep their patients hospitalized longer than do physicians elsewhere. In 2004, doctors kept the mean in-patient stays at 36.3  days in Japan, but at barely 13.4 days in France, 10.4 days in Germany, 7.2 days in the United Kingdom, and 6.5 days in the United States (Nihon iryo seisaku kiko 2007, 242–43). Because many physicians run pharmacies in their clinics, they also profit on the drugs they prescribe (again at government expense). Predictably, observers complain that physicians prescribe too often. Aggregate levels do remain in line with international levels: US patients spent $702 per capita in 2003 on pharmaceuticals, the French $582, the Japanese $509, Germans $468, and Canadians $453 (Nihon iryo seisaku kiko 2007, 217, 247, 253). Yet, because the government negotiates low prices from the pharmaceutical fi rms, the low aggregate figures may disguise large quantities consumed. Of all practicing doctors in Japan, 93,000 work in a small clinic (Kosei rodo sho 2006b, table 2–46). Defi ned as institutions with fewer than twenty beds, these are private affairs. Among these physicians, 71,000 either own the clinic directly or (effectively) own it through a nonprofit organization they control. The other 22,000 work for someone else. Having invested heavily in their clinics, the owners often hope to transfer them to their sons or daughters. Unfortunately, half the children in this world are not as smart as their parents. Given the difficulty their children sometimes have in gaining admission to medical schools, several private schools function—effectively—as medical schools of last resort for the not-very-bright offspring of very wealthy clinic owners. Teikyo University in Tokyo, for example, demands tuition and fees of 14.2 million yen in the fi rst year. Over the six years of medical school education (as of 2013), it collects tuition and fees (excluding room and board) of about 45.2 million. Kawasaki Medical University collects 45.5 million. Most of the remaining physicians work as salaried employees at the larger hospitals (Kosei rodo sho 2006b, tab. 2–46). The most sophisticated doctors work at the hospitals associated with elite medical schools. Others practice at the larger hospitals run by national, prefectural, or municipal governments. Still others work at the hospitals operated by major charitable organizations like the Red Cross. The clinic doctors do not have admitting privileges at these hospitals. Disproportionately, the doctors with private clinics are old. They are not old because—in the language of a Stanford University study— younger physicians are not “interested in clinic-based medicine,” be-

84

Chapter Four

cause “the escalation of land prices in Japan has made the creation of clinics by young doctors fi nancially prohibitive,” or because young doctors “are attracted to . . . sophisticated medical equipment” at the bigger hospitals (Yoshikawa, Bhattacharya, and Vogt 1996b, 27). Instead, the doctors with the clinics are old because of regulation. Nominally to limit hospitalization costs, the Japanese government in 1985 capped the number of hospital beds per medical district. As of 2005, only ten of forty-seven prefectures had fewer beds than were allowed. All others remained beyond their allowed capacity. Tokyo was 6 percent above its cap (106,000 beds with only 100,000 allowed), and Osaka was 16 percent above (Kokuritsu shakai hosho 2006, 428, table 229). In effect, the hospital bed cap bans new construction in most areas. Although the most talented doctors work at university or government hospitals, many will eventually shift into the private sector for the higher incomes. At public-sector hospitals, they simply cannot earn as much as they could at their own institutions. To earn higher incomes, they will eventually need to run their own clinics or hospitals. Individual biographies (see, e.g., Iji koron sha 2004) detail this dynamic. Many clinic owners came to their clinics from a career at a publicsector hospital. Before the ban on new beds, they could build their own clinics. In the years since the ban, in most cities they must buy an existing clinic, build a bedless clinic, or move outside the national insurance system altogether. C. The Political Economy The government sets prices in the health- care industry through negotiations with the physician trade association. Every other year, representatives of the MHLW negotiate a fee schedule with the JMA (Campbell and Ikegami 1998, chap. 6). The JMA, in turn, advances the interests of the clinic doctors. It may include only 61 percent of all Japanese doctors, but it includes virtually everyone who runs a clinic.18 By all accounts, the government sets prices low, but low in a way that favors the clinic doctors over their hospital competitors. According to Campbell and Ikegami (1998, 147), it sets the prices at about one- quarter of the level the service would cost in the United States. Crucially, it skews the prices in ways that divert revenue away from doctors who invest in specialized expertise. It diverts revenue toward those who invest in the small clinics.19

Few Claims, but for a Different Reason

85

As Campbell and Ikegami (1998, 84, 173– 74) explained it, the government “makes inexpensive primary care relatively profitable and expensive high-tech procedures unprofitable.” This “[c]ontinued domination by the JMA” of health policy, Campbell and Ikegami (1998, 174) wrote: [has] left hospital services, especially high-tech medicine and nursing, poorly reimbursed, with no provision for capital investment or administrative overhead. . . . [O]ffice-based physicians and the government have become de facto allies in maintaining the status quo by preventing the encroachment of hospitals and the expensive high-tech medicine that they promote.

Long the single largest donor to the ruling Liberal Democratic Party, 20 the JMA uses its power in ways that extend beyond the price schedule. Through the regulatory structure, it maintains a variety of anticompetitive restraints:21 caps (as noted above) on new beds in a locality, advertising restrictions, higher fees for patients who try to consult with a sophisticated hospital without first visiting a small clinic, 22 and bans on corporate hospital ownership. Informally, when local doctors decide to oppose the construction of new and more sophisticated public hospitals, the government often caves. 23 American physicians lobby for municipal hospitals because they need places to admit their patients. JMA physicians face no such incentive. Instead, they earn the most if they keep their patients out of the hospitals and in their own clinics. To them, a community hospital is simply a higher quality competitor for their customers. They fight plans to build new municipal hospitals in their cities, and the government defers.

D. Consequences 1. not service By skewing pricing away from sophisticated care, the Japanese government neither improves primary care nor promotes “preventive” medicine. Because the government sets even primary care prices below market- clearing levels, doctors relentlessly depreciate quality. First, waits are long. After all, the government limits the number of physicians, subsidizes the cost of medical services, and suppresses the prices paid to doctors. Physicians respond by allocating services by queue. Most doc-

86

Chapter Four

tors take no appointments and treat all patients on a fi rst- come, fi rstserved basis. When patients do make appointments, it avails them little. According to a 2005 MHLW survey, only 45 percent of the patients who arrive at a hospital without appointments see a doctor within thirty minutes, and another 23 percent see a doctor within the next thirty minutes. With an appointment, the fraction able to see a doctor within thirty minutes rises only to 53 percent. 24 At the large high- quality hospitals, patients wait even longer. At the biggest hospitals, only 22 percent of patients without appointments see a physician within thirty minutes. Another 22 percent see one in the next thirty minutes. Thirty-two percent wait one to two hours, and the remaining quarter wait over two hours. 25 Second, visits are short. Again, because the government limits the supply of physicians, subsidizes purchases, and suppresses prices, physicians maximize their incomes by seeing as many patients as possible per day. This strategy, of course, is equivalent to spending as little time with each patient as possible. According to a common Japanese adage, patients “wait 3 hours for a 3-minute consult.” An exaggeration to be sure, the adage nevertheless captures the gist of a Japanese patient’s alltoo-routine plight. To one ambitious internist (Ikegami 2004), the adage was a chance to sell books. The question, as he saw it, was how to make the most of the inevitably short consultation. “How to Be Ten Times Happier with Even the Three Minute Consult,” he titled his book. Inter alia, he offered advice on how to “make the most of an emergency three-minute consult.” More poignantly, he offered advice on “making the most of your child’s three-minute consult,” as well. Third, facilities are poor. To be sure, doctors will invest in equipment if the reimbursement rates pay them a return on their investments. Yet, if they can fi ll their days at the specified rates with poor rather than attractive facilities, they will offer poor facilities. And they do. 2. not preventive care The national health insurance does not promote preventive care for a simple reason: it does not cover it. The program only covers treatments for accidents and disease, and preventive care falls under neither. Some middle- class Japanese do obtain excellent preventive care, but many of them pay for it in cash. Others obtain it as a (not legally required) fringe

Few Claims, but for a Different Reason

87

benefit through their employer. For the popular batteries of annual tests called “human docks,” they pay 40,000–100,000 yen. 26 3. not long lives Japanese live long, but not because of the medical care. At birth, white American males can expect to live seventy-five years (females, eighty). Japanese males can expect to live seventy-nine years (females, eightysix). Even at age forty, Japanese men face a life expectancy of another forty years (women, forty-six) while white American men face only thirty- eight years (women, forty-two). Life expectancy depends on many factors, however, of which sophisticated medical care is but one. Of those factors, it is not even the most important. Clean water, sanitation, and treatments for infectious diseases also matter (Cutler and Miller 2004; Cutler, Deacon, and Lleras-Muney 2006), and on these factors the United States and Japan do not markedly differ. Smoking matters as well—and Japanese do still smoke more than Americans. But food and exercise also matter. Japanese eat less saturated fat and eat less generally. Given urban geography, they walk much farther. As a result, they stand considerably trimmer than most Americans. Among Americans, 34.1 percent are overweight (BMI of 25– 30) and 32.2 percent are obese (BMI over 30). Among Japanese, only 20.3 percent are overweight and barely 3.1 percent obese (WHO 2008). Excess weight takes a large toll. By age forty, an overweight man can expect to live three fewer years. An obese man can expect seven fewer (Peeters et al. 2003). As Comanor, Frech, and Miller (2006, 22; see also Frech 2008) put it, “the relatively poor health outcomes reported for the United States result from a particular risk factor prominent in the U.S.: high obesity rates.” The longer lifespan in Japan than in the United States does not reflect better medical care. For the most part, it simply reflects the choices people make about calories in and calories out. 4. not specialized expertise The universal health insurance does insure that doctors not specialize. Effectively, it eliminates any fi nancial incentive for them to do so. Because the universal coverage boosts demand while the licensing regime cuts supply, Japanese physicians can fi ll their days at government rates. They will fi ll their days at government rates if they spend years acquiring

88

Chapter Four

specialty and subspecialty skills. And they will fi ll their days at government rates if they invest in no specialty training at all. Predictably, most Japanese doctors choose not to acquire specialized expertise. They do what they must for their basic license, but no more. Of the 19,000 JMA members in Tokyo (56 percent of all Tokyo doctors), barely 1,100 advertise themselves as board certified. Earning no returns to specialization, those at the clinics treat (virtually) anyone who walks in the door. A Stanford study declared that “[s]pecialty board certification does not exist in Japan” (Yoshikawa, Bhattacharya, and Vogt 1996b, 27). That overstates the point, but comes remarkably close to the truth. There are indeed boards that certify doctors in Japan, but, because physicians can fi ll their days at government-mandated rates without specializing, few bother. Instead, a majority of Tokyo JMA members in 2004 advertise themselves as specialists in at least two areas and often many more (Ramseyer 2009b). Many advertised services in combinations of fields—like internal medicine and surgery—that most physicians would not consider even remotely related. One physician, for instance, advertised himself as an internist, orthopedic surgeon, cardiologist, dermatologist, allergist, rheumatologist, pulmonologist, gastroenterologist (including endoscopies), and nephrologist—and claimed to practice rehabilitation medicine and physical therapy besides (Iji koron sha 2004, 456). Contrast the United States. A US doctor who chooses family practice will spend a few years in residency and then (after three years in private practice) earn $200,000. Should he specialize more narrowly, he will spend longer training but earn more thereafter. Indeed, to become a heart surgeon he would spend many more years in residency but then make over $500,000. 27 In effect, the doctor invests in his human capital at the outset and earns a return on that investment over the rest of his career (Dranove and Satterwaite 1991, 52). To make credible their claims (to the public) about their expertise, American doctors obtain board certification. Of the approximately 700,000 US physicians, the American Board of Medical Specialties lists 695,000 as board certified. Of these, over 17,000 are board certified in cardiovascular disease and nearly 6,000 in interventional cardiology (both are subspecialties of internal medicine). About 5,700 are board certified in thoracic surgery, 3,600 in neurology, and 3,800 in neurological surgery (American Board of Medical Specialists 2009). By contrast, most Japanese doctors choose not to acquire any spe-

Few Claims, but for a Different Reason

89

cialized expertise. To be sure, certification has increased over the past several years. In 2003, for example, physicians began a certification program in cardiovascular surgery—but only in 2003. By the end of 2005, the program had certified 1,643 Japanese doctors—but of those, it had grandfathered 1,551. Only 92 had actually passed the new standards (Tabayashi 2007, 218–19). 5. not sophisticated procedures Through rates skewed toward primitive care over sophisticated, the Japanese insurance program reduces the number of doctors and hospitals who offer the more sophisticated and complex procedures: bypass operations and angioplasty for heart disease, for example, carotid angioplasty and endarterectomy to prevent strokes, or the complex operations and chemotherapies for cancer. Potentially, these technologically intensive procedures can save lives. Although a few early studies suggested that some brought only modest returns (McClellan, McNeil, and Newhouse 1994), more recent work indicates that—when used appropriately—they can generate large benefits. 28 Japanese doctors perform these complex procedures far less often than do their US peers (as discussed in more detail in section IV.B, below). In 2005, for instance, American doctors performed 469,000 cardiac bypass (coronary artery bypass graft [CABG]) operations and 1.27 million angioplasties. Although Japan had about a quarter the number of deaths from heart disease, Japanese doctors performed less than 3 percent of the US bypass operations (12,000) and less than 6 percent of the angioplasties (70,000–100,000). 29 The Japanese government cuts its cancer treatment costs by refusing to license the new chemotherapy drugs. 30 Pharmaceutical research is expensive. Even as the industry develops ever-more- effective chemotherapy regimes, it has paid for the research with ever-higher prices. When proven effective, the US government has approved these drugs for use. Nominally out of safety concerns, however, many of them the Japanese government has refused to approve for its insurance coverage. Denied access to the most effective treatment regimes, more and more Japanese cancer patients simply abandon the universal insurance. They cannot formally abandon it, of course. But rather than make do with its limited chemotherapy options, they turn to a growing group of oncologists who offer the new (US licensed) treatments on a cash basis.

90

Chapter Four

In part to help Japanese plan for their possible off-universal-insurance chemotherapy needs, an increasing number of insurers offer specifically “cancer insurance.” Aflac was said to dominate the market. In 2007, it alone sold 639,000 new cancer insurance policies. At least eight other fi rms offered the insurance as well. 31 By 2013, the number of fi rms offering the insurance had climbed to fi fteen.

III. One Effect on the Allocation of Talent A. Introduction Consider how this universal insurance might skew the allocation of talent and the development of expertise. The insurance does not cover all procedures. Instead, it excludes those elective procedures thought medically most unnecessary. In effect, it bifurcates the medical services industry into the covered (noncompetitive and medically necessary) sector and the noncovered (competitive and medically superfluous) sector. Within the covered sector, it suppresses prices and subsidizes demand, and physicians respond by degrading quality and allocating service by queue. Within the noncovered sector, physicians allocate service by quality and price. For two reasons, this situation creates an incentive for the best doctors to shift into the medically least necessary sectors and there invest in skills and facilities at higher rates than their peers. First, in the competitive sector, prices vary by physician quality, but in the covered sector they do not. As a result, the most talented physicians could conceivably earn higher returns in the competitive sector through the higher prices they could charge. The less talented would earn more in the subsidized, price- controlled sector. Second, in the competitive sector, physicians can earn a market return on many investments in human and physical capital. If consumers value specialized expertise highly enough, under competition physicians will fi nd it worthwhile to acquire the expertise. If consumers value the certification of that expertise, physicians will fi nd it worthwhile to obtain the certification. And if consumers value attractive facilities, physicians will fi nd it profitable to build high- quality clinics. Three potentially observable implications follow. First, when physicians sell their services in the competitive market, they will offer the talent and the level of expertise, certification, and facilities that they can

Few Claims, but for a Different Reason

91

profitably provide at the price patients are willing to pay. Second, when they sell their services in the market with subsidized demand and suppressed prices, they will present less talent and offer lower levels of investment in human and physical capital. Last, if the most talented doctors shift to the competitive sector and there make greater investments (as the fi rst two hypotheses suggest), then observed incomes should be higher in the competitive (noncovered) sector than they are in the pricecontrolled (covered) sector. To test this logic, I collect biographical information on all 495 Tokyo physicians advertising cosmetic procedures (cosmetic surgeons) and compare them against several other groups of physicians. 32 The data confi rm what the economic logic suggests. Japanese cosmetic surgeons are more talented and better trained than other Japanese physicians— more talented and better trained even than plastic surgeons not advertising cosmetic procedures. They are more likely to have attended one of the more prestigious public medical schools. They are more likely to have served on the faculty of a medical school. And they are more likely to have invested the time necessary to become board certified. Given the talent and expertise they bring, cosmetic surgeons earn high incomes. They earn more than other doctors generally. They earn more than other doctors in their age cohort. And they earn more than noncosmetic plastic surgeons. B. Cosmetic Surgery Crucially, the Japanese national health insurance excludes cosmetic procedures. To be sure, it covers corrective operations for a child burned in a fi re or a passenger disfigured in a car crash. It does not cover cosmetic surgery for otherwise healthy people. Despite this exclusion from the national system, Japanese physicians do sell cosmetic services. According to the MHLW, only 715 physicians practice in the field, and only 239 of them do so in Tokyo (2004 data; Kosei rodo sho 2006b, 183). But the MHLW collects data only on services covered by the national insurance. Among cosmetic surgeons, it counts only those who handle the insured burn treatments, surgeries related to accidents, and other medically necessary procedures. Google “cosmetic surgery,” however, and 449 physicians advertise on the web in Tokyo alone (search in 2007). Google the two professional associations of cosmetic surgeons, and one has 987 members and the other

92

Chapter Four

667 (www.jsaps.com and www.jsas.or.jp, respectively). These doctors offer a wide range of procedures. One of the cheapest and most common involves eyelids. Westerners tend to have a crease in their eyelids, while Japanese do not. Over the past two decades, many Japanese (particularly women) have added that crease through surgery. Cognoscenti describe the operation much the way French chefs describe the omelet— any chef can make one, but only the most skilled and best trained can appreciate the subtleties that distinguish the average omelet from the exquisite. The price for the eyelid operation depends on the physician’s reputation and method, but—like the omelet—it can be cheap. The massive (thirty-four outlets nationwide in 2007) Kanagawa Clinic chain offered the crease for as little as 60,000 yen. Its lead physician, Katsumi Izawa, claimed to have done over 15,000 eyelid operations. His colleague Katsuyuki Yoshitane reported 16,000. According to the websites, most cosmetic surgeons cater to women. Some procedures (e.g., LASIK surgery for nearsightedness) they offer to both sexes, and a few (e.g., penile inserts) they sell only to men. But mostly the clinics target women. For them, they offer face-lifts, nose jobs, liposuction, breast enhancements, hair removal, Botox treatments, sweat-gland removal, and a wide variety of other procedures. Cosmetic surgeons work hard to attract clients, and their websites illustrate how hard. Japanese medical clinics are famously run down; the cosmetic clinics advertise interiors as swank as anything in the Architectural Digest. The typical Japanese doctor either uses no website or posts gruesomely clinical photographs only a fellow physician would think inviting. The cosmetic surgeon hires a professional web designer and lavishly illustrates his site with professional models, lyrical motifs, and soft pastels. With its nearly three dozen clinics in 2007, the Kanagawa Clinic epitomized the successful cosmetic fi rm. It did not succeed by pastels alone. Head of the clinic’s LASIK-surgery practice was Yoshihiro Kitazawa. Formerly on the faculty of the prestigious Tokyo Medical and Dental University, by 1996 he could claim 17,000 LASIK operations. He supervised twenty- eight other physicians at the clinic’s Tokyo LASIK center, and its Osaka branch employed another twenty- one. In the notoriously certification-free Japanese health care industry, all fi fty were boardcertified ophthalmologists. And with its stylish design and medical expertise, the Kanagawa

Few Claims, but for a Different Reason

93

Clinic coupled consumer-friendly fi nance. In the fall of 2004, it offered its usually 60,000-yen eyelid crease at the loss-leader price of 9,800 yen. For all its patients, it offered group discounts, referral discounts, and student discounts. It arranged loans from consumer credit fi rms. It allowed patients to pay on a sixty-payment installment plan. And for its most loyal customers, it offered a “point card system” with which to accumulate credit from one operation toward a discount on the next. C. Data 1. introduction To explore more systematically the effect of universal health insurance on the allocation of talent and the development of expertise, I couple a physician’s biographical information with data on his income tax liability. I collect this material on (1) all physicians who advertise Tokyo-area cosmetic services (about 500 doctors), (2) a random sample of Tokyo members of the JMA (about 500 doctors), (3) all Tokyo JMA members in the same age cohort as the cosmetic surgeons (about 2,200 doctors), (4) all board- certified Tokyo JMA members (about 1,100 doctors); and (5) all Tokyo JMA noncosmetic plastic surgeons (about 100 doctors). 2. income To measure physician incomes, I take from government records the income tax each physician paid in 2004 (Tokyo shoko risaachi 2005). This information is available only for the highest income taxpayers (HIT). The tax amount a person needs to pay to trigger this public disclosure has varied over the years, but in 2004 stood at ten million yen. Because Japanese marginal tax rates peak at 37 percent, 33 a person paying ten million yen in taxes would have made about forty million yen. The 2004 tax year is the last year for which the government made this information available. Under the new privacy statute, the government may not publish the HIT list. 34 In 2004, some 73,000 Japanese paid the ten million yen or more in taxes that landed them on the list. For the physicians in this group, I know the amount of taxes they actually paid. For physicians not on the list, I know that they paid less than ten million yen. As a source of information, tax records inherently present several limitations. First and most obviously, taxpayers have an incentive to

94

Chapter Four

underreport. With a top marginal bracket of 37 percent, the incentive is strong. Although the Japanese tax and prosecutors’ offices punish cheaters severely, my data will still include some doctors who hide income. Second, the amount of underreporting will vary with a physician’s practice. If a doctor bills the national health insurance, the government (as payor) will maintain records on most of the doctor’s revenue. If he offers elective cosmetic surgery, he need never enter that income on the accounting books. As a result, the tax data should reflect relatively accurately the income of most of the doctors in the sectors covered by the national insurance. It may significantly understate the income of the cosmetic surgeons. Last, to the extent that physicians earn returns from their outside investments, their taxable income will overstate their returns from medicine. Because the oldest physicians will have accumulated the greatest wealth, they will also earn the most investment income. As a result, the fraction of taxable income from a medical practice should fall with age. 35 3. cosmetic surgeons Cosmetic surgeons differ from other Tokyo doctors along several dimensions. First, cosmetic surgeons are young. The field itself is new, and the age of its specialists reflects this. According to the MHLW, the number of cosmetic surgeons (practicing within the ambit of the national insurance) doubled from 1984 to 1994 and doubled again by 2004 (Kosei rodo sho 2006b, 46). The data in table 4.3 compare the age distribution of all Tokyo physicians (the information is available in aggregate form in Kosei rodo sho 2006b) with the distribution of cosmetic surgeons. The median male Tokyo physician is aged 45–49 years; the median Tokyo cosmetic surgeon is aged 40–44 years. Seventeen percent of male Tokyo physicians are aged 65 years or older; only 8 percent of the male Tokyo cosmetic surgeons are that old. Second, women are more likely than men to choose to practice cosmetic surgery. As in the United States, women in Japan disproportionately shoulder the burden of childcare and housework. Predictably, they choose specialties with fewer emergency procedures: ophthalmology, dermatology, and anesthesiology (see tables 4.4 and 4.5). At least outside the national insurance system (which covers burn and traffic accident patients), cosmetic surgery similarly involves few emergencies. Among all

Few Claims, but for a Different Reason

95

Table 4.3 Age distribution of Tokyo physicians ALL Age, years ≤24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 ≥85 Total Mean age

Male 41 (.2) 2,448 (9.2) 3,433 (12.9) 3,340 (12.6) 3,663a (13.8) 3,404b (12.8) 2,578 (9.7) 1,978 (7.5) 1,173 (4.4) 1,007 (3.8) 1,046 (3.9) 1,547 (5.8) 639 (2.4) 234 (.9) 26,531 48.2

JMA members Female

38 (.5) 1,645a (20.7) 1,512 (19.1) 1,114b (14.0) 951 (12.0) 725 (9.1) 480 (6.1) 380 (4.8) 218 (2.7) 164 (2.1) 190 (2.4) 292 (3.7) 143 (1.8) 80 (1.0) 7,932 42.6

Male

Female

0 (0) 0 (0) 1 (.2) 10 (2.2) 43 (9.3) 54 (11.6) 52 (11.2) 53 (11.4) 47 b (10.1) 46 (9.9) 54 (11.6) 70 a (15.1) 27 (5.8) 7 (1.5) 464 61.7

0 (0) 0 (0) 4 (5.8) 5 (7.2) 5 (7.2) 10 (14.5) 10 a (14.5) 6b (8.7) 3 (4.3) 5 (7.2) 2 (2.9) 7 (10.1) 9 (13.0) 3 (4.3) 69 58.9

Cosmetic surgeons Male

Female

0 (0) 10 (3.7) 24 (8.8) 32 (11.7) 72 a,b (26.4) 53 (19.4) 28 (10.2) 17 (6.2) 16 (5.8) 9 (3.3) 5 (1.8) 4 (1.5) 3 (1.1) 0 (0) 273 46.4

0 (0) 8 (9.9 15 (18.5) 25a,b (30.9) 16 (19.8) 9 (11.1) 1 (1.2) 2 (2.5) 2 (2.5) 1 (1.2) 0 (0) 2 (2.5) 0 (0) 0 (0) 80 39.4

Sources: Data on all physicians are from Kosei rodo sho (2006) and represent aggregate statistics distributed by the Ministry of Health, Labor, and Welfare; data on randomly sampled Japan Medical Association (JMA) members are from Iji koron sha (2004); the data on cosmetic surgeons are from Iji koron sha (2004) and a World Wide Web search for Tokyo physicians who advertise cosmetic procedures. Note: Data are no. (%) of physicians. a Modal group. b Median group.

Table 4.4 Most common specialties for male physicians in Japan, 2004

Specialty Internal medicine General surgery Orthopedic surgery Pediatrics Psychiatry Gastroenterology Cardiology

Percentage of all male physicians

Percentage male in the specialty

29.2 10.3 8.4 4.7 4.6 4.3 3.8

85.1 95.4 96.3 68.8 81.5 89.5 90.3

Source: Kosei rodo sho (2006).

Tokyo doctors, women are over 60 percent more likely than men to become cosmetic surgeons. Unfortunately, the nature of this project prevents me from assembling more than one year’s data. Because cosmetic surgeons operate outside the national insurance system, the government does not collect data on them. Because membership is voluntary, neither do the professional

Chapter Four

96 Table 4.5 Most common specialties for female physicians in Japan, 2004

Specialty Internal medicine Ophthalmology Pediatrics Dermatology Psychiatry Obstetrics/gynecology Anesthesiology

Percentage of all female physicians

Percentage female in the specialty

26.0 10.9 10.9 7.0 5.3 5.3 4.4

14.9 36.8 31.2 38.0 18.5 21.8 29.1

Source: Kosei rodo sho (2006).

associations. To locate physicians hoping to sell cosmetic surgery, I thus turn to the World Wide Web, on which most cosmetic surgeons advertise. 36 The constantly updated nature of the web, however, prevents me from retrospectively collecting data on physicians who sold cosmetic surgery in the past. Unfortunately, changes in the privacy protection law prevent me from obtaining any income data since 2004. 4. reference groups (a) Introduction For a reference group against which to compare cosmetic surgeons, I turn to the JMA. I do this reluctantly, as the practice introduces several biases. The association does not include all physicians. While Japan has 270,371 licensed physicians, the JMA includes only 164,110 (61 percent). 37 While 34,463 physicians work in Tokyo, the JMA has only 19,418 (56 percent). Crucially, however, microlevel information on all Japanese doctors is simply unavailable. The JMA database introduces at least three biases relevant here. First—and most fundamentally—JMA members are older. While the median male Tokyo physician is aged 45–49, the median male JMA member is 60– 64. While the modal male Tokyo physician is aged 40–44, the modal JMA physician is 75– 79. And while only 17 percent of male Tokyo physicians are 65 or older, 44 percent of the male JMA members are (table 4.3). Second, the JMA disproportionately recruits its members from among those doctors who own and operate their own clinics or hospitals. According to the government, 76,897 (28 percent) of all Japanese physicians operated their own hospitals or clinics in 2007; of all To-

Few Claims, but for a Different Reason

97

kyo physicians, 8,878 (26 percent) do. As the JMA website claimed (in 2007) 84,562 owner- operators in its membership, either the MHLW is undercounting owners or the JMA is overcounting. Either way, most owner- operators must have joined the JMA. Within Tokyo, the JMA claims 9,909 owner- operators. Again, most owner- operators must be in the JMA. Last, JMA members are men. Women attended medical school less commonly decades ago and less frequently own hospitals or clinics today than do men. Since JMA members are older and more likely than most doctors to own their own facilities, they are also more likely to be men. In Tokyo, 77 percent of all physicians are men. In my JMA sample, 87 percent are. (b) Random Sample I fi rst compare the cosmetic surgeons against a random sample of 495 JMA members. To address the greater likelihood that the JMA members own their own clinics (and differ on other omitted variables), as appropriate I limit the cosmetic surgeons in the comparison to those 181 physicians who are also members of the JMA. To address any bias caused by the greater likelihood that the JMA members are men, as appropriate I limit the comparison to men. (c) Age- matched Cohort I then compare cosmetic surgeons against a group of age-matched JMA physicians. As I noted above, JMA members are much older than cosmetic surgeons: only 84 cosmetic surgeons are over age 50, while only 73 JMA members in the random sample who are not cosmetic surgeons are under age 50. To address the problems caused by the largely nonoverlapping ages of the physicians in the two data sets, I collect biographical and tax information on JMA doctors closer in age to the cosmetic surgeons. The cosmetic surgeons at the 25th and 75th age percentiles were born in 1955 and 1967, respectively. As a comparison group for these doctors, I assemble biographical and tax data on all 2,182 Tokyo JMA members born between those years. Again, as appropriate, I limit the comparisons to cosmetic surgeons who are members of the JMA or to those who are men.

Chapter Four

98

(d) Board- certified Physicians Doctors in the JMA are older than other doctors, and older doctors seem less likely than others to have obtained board certification. To explore certification patterns more closely, I collect biographical and tax information on all 1,133 board- certified Tokyo JMA doctors. (e) Plastic Surgeons I assemble biographical and tax information on all 93 noncosmetic Tokyo plastic surgeons in the JMA. Subject to a few exceptions, 38 most cosmetic surgeons are plastic surgeons who perform medically nonessential operations for cosmetic purposes. In Japan, the distinction between cosmetic and plastic surgeons thus tracks the market in which they choose to sell: cosmetic surgeons sell plastic surgery in the uninsured, competitive market, whereas the noncosmetic plastic surgeons sell plastic surgery in the insured, price-regulated market. The similarity between the two groups consequently lets me contrast the qualifications and incomes of otherwise comparable physicians.

D. Talent and Expertise 1. introduction Consider whether talented physicians disproportionately select into cosmetic surgery and, once there, invest at higher levels in specialized expertise. Take talent fi rst and expertise second. I begin with the description of the variables (section D.2), give summary statistics in table 4.6, and turn then to the regression results in table 4.7.

2. variables Tax liability 2004: A physician’s tax liability (in thousands of yen) in 2004 if he is on the list of highest income taxpayers; 10,000 if not on the list. Tax liability 2003: A physician’s tax liability in 2003, conditional on appearing on the HIT list in 2004. HIT: 1 if a doctor appeared on the list of highest income taxpayers in 2004; 0 otherwise.

Few Claims, but for a Different Reason

99

Table 4.6 Tokyo physicians: Selected summary statistics for cosmetic surgeons and JMA members Faculty experience

Public university

Board certification

Cosmetic surgeons Random sample of physiciansa Plastic surgeonsa

.228 (492) .153 (489) .061 (82)

.406 (392) .352 (395) .350 (80)

.432 (491) .099 (486) .160 (81)

.119 (495) .139 (495) .097 (93)

Born 1955–67 Cosmetic surgeons All physiciansa Plastic surgeonsa

.319 (182) .121 (1,369) .065 (31)

.418 (177) .294 (2,079) .294 (34)

.577 (182) .268 (1,367) .161 (31)

.154 (182) .103 (2,099) .088 (34)

Males born 1955–67 Cosmetic surgeons All doctorsa Plastic surgeonsa

.336 (146) .124 (1,170) .091 (22)

.455 (143) .308 (1,820) .333 (24)

.575 (146) .264 (1,168) .182 (22)

.185 (146) .103 (1,839) .042 (24)

Board-certified physicians born 1955–67 Cosmetic surgeons All doctorsa Plastic surgeonsa

.476 (105) .331 (366) .4 (5)

.373 (102) .201 (364) .2 (5)

HIT

.143 (105) .134 (366) 0 (5)

Sources: Data on Japan Medical Association (JMA) members are from Iji koron sha (2004); data on cosmetic surgeons are from Iji koron sha (2004) and a World Wide Web search; and income data are from Tokyo shoko risaachi (2005). Note: Data are mean values, with the number of observations in parentheses. HIT: highest income taxpayers. a Excluding cosmetic surgeons.

Cosmetic surgery: 1 if a physician advertises on the web or in the JMA directory as a Tokyo-area cosmetic surgeon; 0 otherwise. This variable obviously does not capture expertise but does capture the decision to try to sell services in this market. Public university: 1 if a physician attended one of the fi fty public medical schools; 0 otherwise. The public schools are generally more selective than the thirty private medical schools. JMA member: 1 if a physician is a member of the JMA; 0 otherwise. Faculty experience: 1 if a physician has worked on the faculty of a medical school (including service as a research associate [joshu] or an adjunct; 0 otherwise. The variable indicates whether the doctor impressed scholars enough for them to appoint him to the faculty. Hospital physician: 1 if a physician is affi liated with a hospital; 0 otherwise. Board certification: 1 if a physician is a board- certified specialist; 0 otherwise.

Chapter Four

100 Table 4.7 Selection and investment among Tokyo physicians: Probit regressions Board certification 1 Cosmetic surgery Public university Faculty experience Hospital physician Age Age2 Sex N

2

.369*** .242*** (6.22) (5.44) −.033 −.048* (.88) (1.74) .281*** .521*** (5.10) (15.94) −.042 −.038 (.92) (.56) .032** .063 (2.36) (.64) −.0003*** −.0007 (2.67) (.65) −.073 −.037 (1.08) (1.04) 497 1,528

Cosmetic surgery

3

4

.357*** (5.62) −.040 (1.03) .250*** (4.56) −.057 (1.29) .035** (2.20) −.0003** (2.42)

.245*** (4.97) −.053* (1.82) .516*** (4.56) −.040 (.57) .022 (.20) −.0002 (.19)

440

1,296

5

6

.049 .085*** (1.23) (4.25) .173*** .194*** (3.36) (6.09) −.137*** .038 (4.00) (.77) −.016 −.046 (1.17) (.85) .0000 .0003 (.24) (.55) −.026 −.042* (.45) (1.74) 502 1,530

7

8

.160*** .084*** (3.12) (4.06) .151*** .185*** (3.55) (5.50) .320** .016 (2.21) (.34) −.016 −.025 (.12) (.43) −.0002 .0001 (.11) (.19) −.051 (.94) 466 1,298

Sources: Data on Japan Medical Association (JMA) members are from Iji koron sha (2004); data on cosmetic surgeons are from Iji koron sha (2004) and a World Wide Web search; and income data are from Tokyo shoko risaachi (2005). Notes: Regressions give the marginal effect of the variable, calculated at the mean. Absolute values of the z-statistics are in parentheses. All regressions include a constant term. Regression 1 uses those cosmetic surgeons who were members of the JMA and the JMA random sample. Regression 2 uses all physicians in the 1955–67 age-matched data set. Regression 3 uses the same data set as regression1 but limited to men. Regression 4 uses the same data set as regression 2 but limited to men. Regression 5 uses those cosmetic surgeons who were members of the JMA and the JMA random sample. Regression 6 uses all physicians in the 1955–67 age-matched data set. Regression 7 uses all physicians in the 1955–67 age-matched data set who were board certified. Regression 8 uses the same data set as regression 6 but limited to men. *Statistically significant at the 10 percent level **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

Age: The physician’s age, estimated from the year of school graduation if birth year is unavailable. Sex: 1 if a physician is male; 0 otherwise.

3. summary statistics (a) Medical School I offer two indices of physician quality: whether a physician attended a public medical school and whether that physician spent time on the faculty of a university. All else being equal, public Japanese medical schools are more selective and prestigious than private ones. According to one national network of exam-preparation centers, only six private universities (led by Keio and Jikeikai universities) maintain more selective medi-

Few Claims, but for a Different Reason

101

cal schools than the bottom-ranked public university, Gunma University (see Yoyogi zeminaaru 2007). Where 41 percent of the cosmetic surgeons studied medicine at a public university, only 35 percent of the JMA random sample did (table 4.6). Yet many of the private medical schools date from the 1970s, and most of the JMA physicians attended medical school long before that time. Among physicians born between 1955 and 1967, 42 percent of the cosmetic surgeons studied at a public medical school; only 29 percent of the JMA doctors did. And among the male doctors in the 1955– 67 cohort, the contrast is more pronounced still: 46 percent of the cosmetic surgeons attended a public medical school, but only 31 percent of the other male doctors did. The cosmetic surgeons are more talented even than the noncosmetic plastic surgeons. To put the point more graphically, the plastic surgeons who perform eyelid crease surgery are more talented than the plastic surgeons who reconstruct faces after traffic accidents. While 41 percent of the cosmetic surgeons attended a public university, only 35 percent of the noncosmetic plastic surgeons did. While 42 percent of the cosmetic surgeons in the 1955– 67 cohort attended a public school, only 29 percent of the other plastic surgeons in the cohort did. And while 46 percent of the male cosmetic surgeons in that cohort attended a public school, only 31 percent of the other male plastic surgeons in the cohort did. 39 (b) Faculty Experience As in the United States, medical schools in Japan typically recruit their faculty from among their best students. The Tokyo cosmetic surgeons were far more likely to have spent time on medical school faculties than were the others: 23 percent of the cosmetic surgeons had spent time on a university faculty; only 15 percent of the randomly sampled JMA members had done so. Among physicians born in 1955– 67, the contrast in faculty experience is greater. Of the cosmetic surgeons, 32 percent had served on a faculty, but only 12 percent of the noncosmetic JMA members. Of the men in this age cohort, 34 percent of the cosmetic surgeons had served on a faculty, but only 12 percent of the others. And of the board- certified physicians in the age cohort, 48 percent of the cosmetic surgeons had served on a faculty, but only 33 percent of the others.

102

Chapter Four

The differences persist even when noncosmetic plastic surgeons are used in the comparison. Of plastic surgeons not advertising cosmetic services, only 6 percent had faculty experience (in contrast to 23 percent of the cosmetic surgeons). Of those in the 1955– 67 birth-year cohort, only 7  percent had faculty experience (in contrast to 32 percent of the cosmetic surgeons). And of the men in that cohort, only 9 percent had faculty experience (in contrast to 34 percent of the cosmetic surgeons).40 (c) Board Certification Cosmetic surgeons are much more likely than other doctors to obtain board certification.41 Most Japanese doctors never pursue boardcertified expertise. Able to fi ll their days with patients without certification, they do not bother. Cosmetic surgeons, however, sell their services in a market where price depends on perceived value, where informational asymmetries plague claims about expertise, and where the patient pays the entire cost of any procedure. In that environment, cosmetic surgeons apparently fi nd it profitable both to invest in specialized expertise and to obtain its third-party certification. Among the randomly sampled JMA members, only 10 percent were board certified; among cosmetic surgeons, 43 percent were. Among the 1955– 67 birth-year cohort, 27 percent of the noncosmetic JMA members were board certified, but 58 percent of the cosmetic surgeons were. And among the men in that age cohort, 26 percent of the noncosmetic JMA members were board certified, but 58 percent of the cosmetic surgeons were. Again, the contrast continues even between the cosmetic surgeons and the noncosmetic plastic surgeons. Only 16 percent of all noncosmetic plastic surgeons were board certified, whereas 43 percent of the cosmetic surgeons were. Only 16 percent of those in the 1955– 67 birth-year cohort were certified, while 58 percent of the cosmetic surgeons were. And only 18 percent of the men in that age cohort of noncosmetic plastic surgeons were certified, whereas 58 percent of the cosmetic surgeons were.42 The typical cosmetic surgeon obtained certification in general plastic (rather than specifically cosmetic) surgery. Within cosmetic surgery itself, two rival professional associations also certify expertise: the Japan Society of Aesthetic Plastic Surgery and the Japan Society of Aesthetic Surgery. Perhaps because of problems stemming from the com-

Few Claims, but for a Different Reason

103

petition between the two groups, or perhaps because certification from an older and better- established group provides greater credibility, cosmetic surgeons tend to turn to plastic surgery for their certification. Others in cosmetic medicine obtain certification in fields like dermatology or ophthalmology. 3. regression results In table 4.7, I regress board certification on a variety of physician characteristics. The coefficient on cosmetic surgery is overwhelmingly statistically significant: cosmetic surgeons are substantially more likely than others to obtain board certification. Those who served on a university faculty are similarly likely to obtain certification. These results hold true whether I compare JMA cosmetic surgeons against the randomly selected doctors (regression 1), whether I compare all cosmetic surgeons born between 1955 and 1967 against JMA doctors in the same age cohort (regression 2), and whether I limit the data to men (regressions 3 and 4). In table 4.7, I also regress entry into cosmetic surgery on physician characteristics. Because the decision to obtain board certification is endogenous to the decision to practice in cosmetic surgery, I omit the certification variable. As the summary statistics in table 4.6 suggest, those who attended a public medical school and who served on a university faculty are more likely than other doctors to become cosmetic surgeons: the coefficient on public university is strongly statistically significant in three of the four regressions, and the coefficient on faculty experience is strongly statistically significant in all. Note that this fi nding holds true even when I restrict the database to board- certified physicians (regression 7). E. Income 1. introduction By the logic above, talented doctors choose cosmetic surgery because of the opportunity to earn a fi nancial return on their higher-than-average talent. Once in the field, they invest in specialized expertise, again because of the chance to earn a return on that expertise. Earning returns that other physicians do not, they should report higher incomes. They do.

104

Chapter Four

2. summary statistics Some cosmetic surgeons are indeed rich. Of the more than 4,000 Tokyo doctors in the pooled data set, seven of the ten with the highest incomes were cosmetic surgeons (table 4.8). These high incomes are not an aberration. Eight of the 2004 top ten (five of the seven cosmetic surgeons) earned enough in 2003 to appear on the HIT list (that is, make forty million yen or more) that year too, and most of them placed high on the 2003 list. Indeed, five of the top ten (three of them cosmetic surgeons, and none of them older than 45) had already appeared on the HIT list at least four other times. Not only do the most successful cosmetic surgeons dominate the topten income list but cosmetic surgeons in general appear on the HIT list in large numbers. Granted, according to the data presented in table 4.6, 14 percent of the 495 randomly sampled JMA members appear on the list, but only 12 percent of all Tokyo cosmetic surgeons. Recall, however, that the JMA members are much older than the cosmetic surgeons. Of those cosmetic surgeons in the JMA, 20 percent appear on the HIT list. Of physicians born between 1955 and 1967, 15 percent of the cosmetic surgeons make the list, but only 10 percent of the others. Of the men born between 1955 and 1967, 19 percent of the cosmetic surgeons appear on the list, but only 10 percent of the others. Specialists earn more than generalists in the United States, too, of course, but Japanese cosmetic surgeons also earn more than their specialist peers. Take the plastic surgeons who choose not to advertise cosmetic services. Whereas 12 percent of the JMA cosmetic surgeons appear on the HIT list, only 10 percent of the other plastic surgeons do (table 4.6). Within the 1955– 67 birth-year cohort, 15 percent of the cosmetic surgeons appear on the list, but only 9 percent of the other plastic surgeons. And among the men in that cohort, 19 percent of the cosmetic surgeons appear on the list, but only 4 percent of the other plastic surgeons.43 3. regression results In table 4.9, I confi rm this logic more systematically. For these regressions, I take as my dependent variable a physician’s tax liability in 2004. I use a Tobit model, since I know only that for a majority of doctors they paid less than ten million yen.44

10 8

1 3 2 31 4 14

Physician rank, 2003 102 225 281 346 1,366 1,584 2,019 2,405 2,704 2,842

Taxpayer rank, 2004 Cosmetic surgery Cosmetic surgery Ophthalmology Cosmetic surgery Cosmetic surgery Cosmetic surgery Cosmetic surgery Cosmetic surgery Internal medicine Surgery

Primary field 313,796 203,986 174,493 159,647 78,832 73,558 65,723 59,730 56,449 54,916

2004 Tax (1000s of yen)

47 44

70 45 42 41 45 46 35

Age, years

Teikyo Nihon

Keio Kagoshima U Tokyo Jikei Yamanashi Showa Kyorin

Medical school

Yes No No Yes No No No Yes No No

Board certified

Sources: Data on Japan Medical Association members are from Iji koron sha (2004); data on cosmetic surgeons are from Iji koron sha (2004) and a World Wide Web search; and income data are from Tokyo shoko risaachi (2005). a Information not available for M. Boku.

1 2 3 4 5 6 7 8 9 10

Physician rank, 2004

Table 4.8 High-income physicians: Pooled sample

497

11,883* 18.5 2,615 5,216 −9,952 751 −12.0 12,481

1

(1.76) (.00) (.42) (.82) (1.41) (.43) (.82) (1.38) 1,528

9,505** −3,564 1,374 −7,485 −3,719 14,578 −160 4,821

2 (2.55) (1.29) (.38) (.99) (1.27) (1.52) (1.47) (1.39) 1,296

13,090*** −4,379 1,669 −6,580 –3,996 19,352* −217*

3 (3.19) (1.44) (.42) (.82) (1.22) (1.80) (1.77)

1,431

20,789*** −3,512 3,038 −5,764 −3,936 13,573 −151 3,299

4 (4.64) (1.24) (.81) (.75) (1.32) (1.37) (1.35) (.95) 1,528

7,541* −1,836 5,948 −1,590 −6,838** 17,309* −187 1,539

5 (1.91) (.64) (1.58) (.22) (2.17) (1.67) (1.59) (.44)

(.12) (1.36) 5,106 (.32) 51,938 (.90) −588 (.89) 58,849* (1.91) 48,382*** (2.78) 177

1,839 −24,450

6

Sources: Data on Japan Medical Association (JMA) members are from Iji koron sha (2004); data on cosmetic surgeons are from Iji koron sha (2004) and a World Wide Web search; income data are from Tokyo shoko risaachi (2005). Note: Regressions give the marginal effect of the variable calculated at the mean. Absolute values of the t-statistics are in parentheses. For all regressions except regression 5, the dependent variable is Tax Liability 2004; for regression 5, the dependent variable is Tax Liability 2003. All regressions include a constant term. Regression 1 compares JMA cosmetic surgeons against a randomly sampled population of the JMA. Regressions 2–6 are limited to physicians born 1955–67. Regression 3 is further limited to males, regression 4 to JMA members, and regression 6 to cosmetic surgeons. Note that in regression 6, Tobit would not calculate the standard errors for Hospital Physician. Results of instrumental variable specifications (instrumenting Cosmetic surgery with Age, Age2, and Sex) are reported in Ramseyer (2007, table 5). *Statistically significant at the 10 percent level **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

Cosmetic surgery Public university Faculty experience Hospital physician Board certification Age Age2 Sex JMA member N

Variable

Table 4.9 Determinants of income for Tokyo physicians: Tobit regressions

Few Claims, but for a Different Reason

107

I begin with the contrast to the random sample of JMA physicians (regression 1). Because JMA doctors differ from others along several unobserved dimensions (most importantly, ownership of a clinic), I limit cosmetic surgeons to those in the JMA.45 Among JMA members, cosmetic surgeons do indeed pay higher taxes than the others do—about twelve million yen more. The results for the other variables are not statistically significant, although men probably earn more than women. As I discussed earlier, however, the randomly sampled JMA doctors are much older than the cosmetic surgeons. Suppose I limit the data set to all physicians born between 1955 and 1967 (regression 2). The statistical significance of the coefficient on cosmetic surgery dramatically increases. If I limit the data to men born during this period, both the statistical significance and the economic magnitude of the coefficient increase further still (regression 3). As one would expect, age and its squared term indicate that incomes increase with age but at a declining rate. If I limit the cosmetic surgeons to JMA members (as I did in regression 1) but maintain the 1955– 67 birth-year restriction, the contrasts between cosmetic surgeons and the others again increase (regression 4). And even if I use tax liability in 2003 rather than in 2004 as the dependent variable, cosmetic surgeons retain their advantage (regression 5).46 Members of the JMA are much more likely than other physicians to own their own clinics. In turn, the taxable incomes of those who own their own clinics should include a return on the investments (both tangible and intangible) that they have made in those establishments. Because they earn returns on investments that others have not made, they should earn higher incomes. In regression 6, I compare the incomes of cosmetic surgeons in the JMA against those not in the association. Among cosmetic surgeons born in 1955– 67, JMA members do indeed earn higher incomes than others do. The largest puzzle in the table 4.9 results concerns the apparent absence of any premium to board certification. Instead, the coefficient on board certification is generally negative, and in one regression statistically significantly so. By logic, whether in a cosmetic field, physicians should invest in specialized expertise only if they earn a return on this investment. Because they earn a return that others do not, they should report higher incomes. Recall, however, that, although no clinic ownership variable appears in the data, half the JMA members own their clinics. What is more,

108

Chapter Four

given the government restrictions on new clinics (described above) and the stratospheric Japanese estate taxes on cash bequests, these clinic owners often transfer their clinics to their children before death. Plausibly, doctors who acquire clinics while young may choose to focus their efforts on learning to run the sometimes-substantial business establishments. By contrast, the doctors who choose to become board- certified may also be those without personal clinics. They may invest in their specialized expertise to compete against their peers with established clinics. If all this is true, then a regression of income on board certification that omits a clinic- ownership variable could easily yield no statistically significant coefficient on certification.

F. Implications 1. contrast with the united states Cosmetic surgeons earn high incomes in the United States as well, of course. They invest years in specialized training and earn a return that reflects in part the returns on that human capital investment (see, e.g., Dranove and Satterthwaite 1991, 52). So, too, in Japan. Talented doctors enter the field, they acquire specialized skills, they certify those skills, and they earn a return that reflects their superior talent and expertise. But the United States and Japan do differ, and the contrast appears not among cosmetic surgeons; instead, it appears among the other physicians. While cosmetic surgeons invest in board- certified expertise, other physicians do not. Even other physicians in plastic surgery—a field demanding the same skills—do not. Cosmetic surgery attracts talented physicians. But, again, plastic surgery—which presents the same intellectual challenges—does not. In both countries, cosmetic surgeons earn incomes that reflect the talent and investments they bring to the job. Yet US physicians in other fields also invest heavily in specialized expertise, while their Japanese peers invest much less. Predictably, given these conditions, although cosmetic surgeons earn high incomes in both countries, physicians in other fields earn relative incomes in Japan that bear no relation to their relative incomes in the United States.47

Few Claims, but for a Different Reason

109

2. selection out of medicine In fact, these data understate the extent to which the price controls distort career choices. Logically, the controls will not skew just a physician’s choice of specialty: they will skew a talented student’s more basic career choice. More specifically, they will steer the most talented science students away from any career in medicine. Speculative (obviously, only speculative) evidence consistent with this broader effect appears in the relative attractiveness of engineering and medicine. Forty-seven universities maintain both engineering and medical schools. The seven worst universities have bad schools in both fields. But among the other forty, the relative attractiveness of medicine falls as the quality of the university rises. Consider the correlation among the forty universities between (1) the selectivity of the medical school and (2) the ratio of the selectivity of the university’s medical school to the selectivity of the university’s engineering school (with selectivity measured by Yoyogi zeminaaru 2007). The coefficient is strongly negative (−.60) and statistically significant at more than the 99 percent level. The better the students at a university, the less attractive a medical career becomes compared to engineering. 3. welfare losses Do Japanese consumers lose when their best doctors migrate into cosmetic surgery? Perhaps cosmetic work yields large returns to unusually high levels of judgment, creativity, and expertise. Perhaps consumers perceive these large differences between high- and moderate- quality cosmetic surgery. And perhaps consumers value the highest quality work highly enough to pay the prices necessary to bring talented doctors into the field. Unfortunately, we do not know—and we do not know because the other sectors of the Japanese health care industry generate no comparable price signals. Consumers may value the difference between highand moderate- quality work in neurosurgery too, but the universal insurance prevents them from bidding for the best surgeons. They may or may not want their best doctors in cosmetic surgery rather than in neurosurgery, but under the current system we have no way to know. And yet we have reason at least to suspect that Japanese consumers suffer a substantial welfare loss. To be sure, even if consumers did pre-

110

Chapter Four

fer to see their doctors’ medical talent in neurosurgery rather than in cosmetic surgery, they would not incur a large loss from the migration of five hundred talented doctors into the latter. Instead, they lose from the failure of the other thirty-four thousand Tokyo doctors to sort themselves among the remaining fields by talent. In other words, consumers may well want their best doctors in some fields rather than others, but the universal health insurance prevents them from sending the price signals necessary to cause that sorting. Similarly, consumers do not incur a large loss from having their five hundred cosmetic surgeons invest in specialized expertise. Instead, they lose from the failure of the other thirty-four thousand to make analogous investments. In other words, consumers may well want their cardiologists and oncologists to invest in specialized skills too, but the current system prevents them from generating the price signals necessary to induce the investments.

IV. One Effect on Mortality A. Introduction By suppressing the development of sophisticated expertise, the national health insurance necessarily shifts the type and level of medical care away from the competitive market equilibrium; in the process, it affects mortality patterns. Section III explores some of the ways it affects the choices talented people make in their careers. Turn now to some of the measurable effects that these choices can have on mortality rates. The national insurance reduces the incentive that physicians face to learn complex, cutting- edge medical care. The insurance keeps costs low by capping the amounts it pays these doctors for the most modern and sophisticated procedures. Facing stringent caps on these procedures, Japanese physicians have little incentive to invest in the necessary expertise. Instead, they invest in small private clinics. In turn, the proliferation of the primitive clinics reduces both the number of complex modern medical procedures performed, as well as the number of doctors and physicians with any substantial experience in those procedures. Yet, for complex modern procedures, that experience is crucial. At least initially, the quantity of a doctor’s and hospital’s experience in a procedure correlates positively with clinical success. Without that accumulated expertise, the mortality rates in the associated dis-

Few Claims, but for a Different Reason

111

eases necessarily climb. Unfortunately for the critically ill patient, Japan does not do many complex procedures—and does not have many institutions that could do them well if they tried. To estimate the costs of this regulatory- driven lack of expertise, I combine mortality data from 855 cities with information on local hospital expertise. In the typical city, I fi nd that the addition of one hospital with substantial experience in modern stroke treatment would cut annual stroke mortality (from the current mean of ninety) by seven to sixteen deaths. The addition of one hospital with substantial experience in angioplasties would cut the annual deaths from heart attacks in the city (from the current mean of fi fty-three) by over nineteen. B. Modern Death 1. causes When they eventually die, both Japanese and US adults die principally of heart disease, strokes, and cancer—though the mix varies a bit across the two countries (see table 4.10). Americans die primarily of heart disease (222 deaths per 100,000), cancer (189 per 100,000), or strokes (51 per 100,000). Among the cancers, they most commonly die of lung cancer (54.2 per 100,000), colorectal cancer (22.2 per 100,000), or breast cancer (15.6 per 100,000). Japanese die of cancers (258 per 100,000), heart disease (137 per 100,000), or strokes (105 per 100,000). Among the cancers, they most often die of lung cancer (44.6 per 100,000), stomach cancer (39.9 per 100,000), or colorectal cancer (30.8 per 100,000).48 2. medical care Because the widespread use of clinics dissipates complex illnesses among nearly a hundred thousand small institutions, it cuts the number of doctors and hospitals with experience in the modern treatment of these diseases—treatments like bypass operations, angioplasty, endarterectomy, or the latest chemotherapies. These technologically intensive procedures save lives. As discussed above, recent work indicates that many generate large benefits and lie at the heart of the procedures that the US medical system provides so extraordinarily well (Cutler and Mas 2004). Doctors and hospitals perform these complex procedures best if they perform them often. Obviously, the amount of experience necessary depends on the procedure, and the marginal benefit of additional experi-

Chapter Four

112 Table 4.10 Selected US-Japan comparisons

Physicians,a n Physicians, per 1,000 Healthcare spending,b amount Healthcare spending, % GDP Healthcare spending, per capita Population aged >65 years,c % Cancer deaths,d n Cancer deaths, per 100,000 Heart disease deaths, n Heart disease deaths, per 100,000 Stroke deaths, n Stroke deaths, per 100,000 Mean in-patient hospital stay, a days Japane

Japan

United States

259,336 2.0 $343.8B 8.0 $2,694 20.1 325,941 258.3 173,125 137.2 132,874 105.3 36.3

700,287 2.4 $1,660.7B 15.2 $5,711 12.4 553,888 188 652,486 222.2 150,004 51.1 6.5

USf (All)

US (White)

Male Female Male Female

Male

Female

Male

Female

Male

Female

Life expectancy, at birth Life expectancy, at 40 years

78.6 39.9

85.5 46.4

74.7 37.2

80.0 41.5

75.3 37.6

80.4 41.8

a

Nihon iryo seisaku kiko (2007) (2004 data). Ibid. (2003 data). Kosei rodo sho (2007; table 1.7, 2005 data); United States, Age Distribution, 2000, www.censusscope.org/US/chart _age.html. d Kosei rodo sho (2007; table 1.26, 2004 data); Heron (2007, tabs 1, C, 2004 data). e Kosei rodo sho (2008a). f Arias (2006; table A, 2003 data). b c

ence will decline as volume increases (Kawabuchi and Sugihara 2006). Repeatedly, however, studies show that a hospital needs a reservoir of experience to perform these complex procedures effectively.49 Japanese hospitals and doctors perform many of the complex procedures far less often than do their US peers. As noted earlier, for instance, in 2005 US doctors performed 469,000 CABG operations and 1.27 million angioplasties, whereas Japan (with a quarter of the deaths from heart disease; see table 4.10) had less than 3 percent of the CABG operations and 6 percent of the angioplasties. 50 Of the 855 cities in Japan (data set described in section C.1, below), 77 percent lacked any hospital with substantial experience in cerebrovascular disease, and 89 percent lacked any with substantial experience in heart disease (with experience defi ned in section D.1, below). The testable implication follows: because accumulated hospital experience improves treatment outcomes, mortality rates from heart and

Few Claims, but for a Different Reason

113

cerebrovascular disease should be highest in areas that lack hospitals with much experience in treating these diseases. After all, as Cutler (2007, 1089; see also Guadagnoli et al. 2000) put it for the United States, “patients are generally taken to the nearest hospital after an MA [heart attack], and where one is admitted affects what is done—even several months later.” The principle that patients do best in areas with sophisticated hospitals clearly applies in the United States, of course (Cutler 2007), but also in Japan. Put most simply, people are substantially more likely to die from heart and cerebrovascular disease in areas without sophisticated hospitals than in areas with them. Unfortunately, the small clinics and hospitals crowd out these sophisticated institutions in most of Japan. 51

C. Data 1. sources To explore these issues, I gather municipal-level data. The forty-seven prefectures in Japan include both cities and other areas. I treat each city as an observation, and the other areas in each prefecture as an additional observation. Because Tokyo is so large (and technically not a city), I treat each ward within Tokyo as a distinct observation. Through this procedure, I create a database of 855 observations. Most of the cities are modest, with the 25th– 75th percentiles ranging from 45,000 to 151,000 by population (table 4.11). The smallest unit is a town in the northern island of Hokkaido; the largest is the city of Yokohama. For each locality, I obtain demographic data from the Ministry of Internal Affairs and Communications (www.stat.go.jp; Somu sho 2007). I also turn to this source for most of the local infrastructural variables used as instruments in the two-stage estimates (discussed below). I take information on the number of deaths and on life expectancy from the MHLW (Kosei rodo sho 2006a, 2006c). The death rates from strokes range from 167 per million population (in Ritto, Shiga prefecture) to 1,899 (in Yuzawa, Akita prefecture; see table 4.11). The death rates from heart attacks range from 90 per million (Urayasu, Chiba prefecture) to 1,342 (Mikasa, Hokkaido). Ritto and Urayasu are near the metropolitan centers of Kyoto and Tokyo, respectively; Yuzawa and Mikasa are in distant northern prefectures. I obtain data on hospital experience from two private sources. The

Chapter Four

114 Table 4.11 Summary statistics

Min Dependent variables Strokes, no. Strokes, per capitaa Heart attacks, no. Heart attacks, per capitaa Hospital expertise variables Stroke hospitals, no. Stroke hospitals, per capitaa Heart attack hospital, no. Heart attack hospital, per capitaa Control variables Population, n Fraction aged 15–64 years Fraction aged ≥65 years Density Instruments Medical schools Business establishments Post offices Parks Libraries

25th percentile

Median

75th percentile

Max

6 167 3 90

32 476 18 268

50 665 29 369

97 902 56 523

1,524 1,899 1,218 1,342

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

14 45 2 23.7

5,221 0.52 0.09 17.0

45,499 0.60 0.18 178.6

74,294 0.64 0.21 439.3

151,240 0.67 0.26 1480.0

3,579,628 0.7 0.4 19,924.8

0 237 3 0 0

0 2,090 11 15 1

0 3,338 17 38 2

0 6,824 31 120 4

4 20,1462 626 2,595 70

Sources: Kosei rodo sho (2006a, 2006c); Somu sho (2007); Shobunsha (2007); Yomiuri (2008); Shukan Asahi (2008). a Numbers given are × 106.

Japanese government requires medical institutions to report the number of operations they do for specified procedures. The Shukan Asahi (2008) magazine obtained this information from local government offices and used it to rank those with more than fi fty cerebrovascular procedures. The Yomiuri (2008) surveyed medical institutions to obtain analogous data. I use the Shukan Asahi for hospital experience with cerebrovascular disease, and the Yomiuri for experience with heart disease. 2. diseases I focus on strokes (the principal cause of death from cerebrovascular disease) and heart attacks (after heart failure, the principal cause of death from heart disease). 52 Cancers present an obvious alternative. After all, they kill a large fraction of the Japanese population but respond to sophisticated modern medical regimes. I use strokes and heart attacks instead because an increasing number of wealthier Japanese have effectively opted out of the national insurance altogether for their cancer treatments

Few Claims, but for a Different Reason

115

(as noted in section II.D.5, above). I know of no analogous trend toward off-insurance treatments for cerebrovascular or heart disease.

D. Variables 1. hospital expertise (a) Stroke Hospitals The Shukan Asahi magazine ranked hospitals by the number of stroke victims to whom the hospitals administered tissue plasminogen activator (tPA) in 2006. As a stroke treatment, tPA is a relatively recent innovation. Though not generally indicated (even the hospitals using tPA most often still gave it to less than 2 percent of their stroke patients), if administered soon after a stroke it can sometimes be very effective. I use the administration of tPA as an index of the extent to which a hospital follows modern cerebrovascular procedures. My stroke hospitals’ variable gives the number of hospitals to have used tPA in the past year in each locality. The number ranges up to 14 (for Osaka); the modal value is 0 (see table 4.11). 53

(b) Heart-Attack Hospitals The Yomiuri magazine ranked hospitals by the number of angioplasties. My heart-attack hospitals’ variable gives the number of top fi fty such hospitals in each locality. The number ranges up to 2 (for five cities); the modal value is 0 (see table 4.11). The most experienced hospital performed 2,393 angioplasties; the fi ftieth hospital performed 583. This is slightly (though only slightly) higher than the cutoff used in prominent studies of the US health- care system. Jollis et al. (1994), for example, categorizes a hospital doing two hundred to four hundred angioplasties as a low-volume (and high-mortality) hospital. Ho (2000) classifies a hospital as high volume (low mortality) if it performed more than four hundred angioplasties per year. 54 2. dependent variables For my dependent variable, I take the numbers of stroke and heart attack deaths in each locality from Kosei rodo sho (2006a). By dividing the numbers by the local population, I create:

Chapter Four

116

Strokes per cap: Number of deaths per capita during 2006 in a locality from strokes (cerebral infarctions). Infarctions can be caused either by ischemia or by hemorrhage. Heart attacks per cap: Number of deaths per capita during 2006 in a locality from heart attacks (acute myocardial infarctions).

3. demographic controls Population: Number of residents in the locality, as of 2005. Fraction 15– 64: Fraction of residents aged 15– 64 in the locality, as of 2005. Fraction 65 and over: Fraction of residents aged ≥65 in the locality, as of 2005. Fraction over 65 squared: Fraction of residents aged ≥65 years squared. Density: Number of residents, divided by the area of the municipality (square kilometers). More fi nely partitioned age data are available at the prefectural but not at the municipal level.

4. instruments Potentially, the number of sophisticated hospitals in an area could reflect the incidence of the disease in the area. To address this possible endogeneity, in regressions 4 and 6 of tables 4.13 and 4.14, I instrument the number of sophisticated hospitals by several variables. The most obvious of these instruments is the number of medical schools in an area. Universities run many of the most sophisticated hospitals and (to my knowledge) do not deliberately locate their teaching hospitals in areas with high death rates from strokes and heart attacks. The other instruments reflect the willingness of a given municipality to invest more generally in modern public services and infrastructure. Medical schools: Number of medical schools in an area, in 2007 (Shobunsha 2007). Business establishments: Number of business establishments in an area, as of 2006 (Somu sho 2007).

Few Claims, but for a Different Reason

117

Post offices: Number of post offices in an area, in 2005 (Somu sho 2007). Parks: Number of parks in an area, in 2005 (Somu sho 2007). Libraries: Number of public library branches in an area, in 2005 (Somu sho 2007).

These are not weak instruments. The fi rst-stage F statistics in all the two-stage estimates in tables 4.13 and 4.14 are substantially above 10, as Stock and Watson (2003, 349– 50) require. The fi rst-stage R 2s are above 0.7 for the stroke regressions and above 0.25 for the heart attack regressions. Given that the best hospitals tend to be university hospitals, the medical schools’ variable is particularly strong: the fi rst-stage t statistic on the variable runs from 7 to 11.8.

E. Results 1. simple contrasts Before turning to the regressions, ask whether people in areas with specialist hospitals enjoy lower mortality rates from strokes and heart attacks (table 4.12, panel A). They do. To measure the difference, partition the cities by whether they have at least one sophisticated stroke hospital or at least one sophisticated heart-attack hospital (as defi ned above). The results are straightforward: areas with a stroke hospital have lower stroke-related mortality rates, and areas with a heart-attack hospital have lower heart-attack–related mortality. Consider whether people in areas with sophisticated hospitals have longer life expectancies (table 4.12, panel B). Again, they do—though with a puzzling sex-related distinction. Men are more likely to die than women of heart attacks (a mortality rate of 450 per million for men and 363 per million for women), but the benefit to men of hospital sophistication seems to come from strokes. At four of the five ages measured, men live statistically significantly longer in areas with sophisticated stroke hospitals than in areas without. For heart-attack hospitals, they enjoy a statistically significant advantage only at age eighty. By contrast, women are more likely than men to die of strokes (a mortality rate of 733 for women and 672 for men), but the benefit from hospital sophistication seems to appear in heart attacks. Again, at every age measured, women live statistically significantly longer in areas with so-

Chapter Four

118 Table 4.12 Regional health effects, by hospital experience A. Regional mortality rates (×10 6), by hospital experience Stroke hospital 602 Heart attack hospital 268

Stroke rate Heart attack rate

No stroke hospital 746 No heart attack hospital 420

Sig. level 0.000 Sig. level 0.000

B. Regional life expectancy, by hospital experience Stroke hospital

1. Men At birth At 20 At 40 At 65 At 80 2. Women At birth At 20 At 40 At 65 At 80

Heart attack hospital

Yes

No

Sig. level

Yes

No

Sig. level

77.843 58.435 39.234 17.577 7.957

77.591 58.211 39.080 17.495 7.910

0.001 0.002 0.020 0.064 0.149

77.636 58.241 39.093 17.570 8.002

77.660 58.274 39.125 17.508 7.910

0.806 0.723 0.703 0.277 0.028

84.619 65.092 45.529 22.413 10.651

84.595 65.082 45.544 22.466 10.655

0.690 0.867 0.779 0.296 0.927

84.772 65.260 45.695 22.607 10.805

84.574 65.057 45.516 22.427 10.829

0.008 0.005 0.012 0.005 0.002

Sources: Kosei rodo sho (2006a, 2006c); Somu sho (2007); Shobunsha (2007); Yomiuri (2008); Shukan Asahi (2008). Notes: Hospital experience variables are as defined in the text. Significance levels are by a two-tailed test. Number of jurisdictions in panel B differs from that in panel A, because life expectancy data are not available for all places.

phisticated heart-attack hospitals than in areas without. For stroke hospitals, the differences are insignificant. These table 4.12 comparisons do not hold demographic composition constant, of course. Yet, strokes and heart attacks are conditions whose incidence increases with age. To hold constant the impact of those demographics, turn now to the regressions in tables 4.13 and 4.14. 2. regressions (a) Introduction To study more closely the effect of hospital experience on mortality, turn in tables 4.13 (for stroke deaths) and 4.14 (for heart attack deaths) to a series of multiple regressions. In these specifications, I regress mortality patterns on the hospital experience and local demographic variables. In both tables, I focus on regressions 1–4. In each of these specifications, I use the per- capita death rate as the

Few Claims, but for a Different Reason

119

Table 4.13 Hospital experience and mortality: Strokes Dependent variable: Stroke deaths 1 OLS (per cap) Stroke hospital Per capita Dummy Number Populationa Fraction 15–64 Fraction ≥65 Fraction >652 Densityb N Adjusted R 2

2 OLS (per cap)

3 OLS (per cap)

4 2SLS (per cap)

5 OLS (no.)

6 2SLS (no.)

−2.557** (2.08) −0.00003* (1.93)* −0.00002** (2.27) 17.5 15.5 55.8 (0.62) (0.01) (1.28) 0.001* 0.001* 0.001* (1.78) (1.80) (1.81) 0.008*** 0.008*** 0.008*** (7.62) (7.58) (7.61) −0.007*** −0.007*** −0.007*** (3.37) (3.34) (3.33) −10.5*** −10.5*** −10.7*** (3.91) (3.92) (3.99) 855 855 855 0.59 0.59 0.59

−0.00003* (1.83) 89.4 (1.36) 0.001* (1.86) 0.008*** (7.64) −0.007*** (3.38) −10.8*** (4.03) 855 0.59

−7.696*** −15.994*** (4.18) (4.80) 0.0005*** 0.0006*** (62.19) (42.95) 144 175 (1.34) (1.60) 1824*** 1894*** (9.07) (9.24) −3084*** −3183*** (8.05) (8.18) −0.002*** −0.002*** (4.29) (4.48) 855 855 0.91 0.91

Sources: Kosei rodo sho (2006a, 2006c); Somu sho (2007); Shobunsha (2007); Yomiuri (2008); Shukan Asahi (2008). Notes: Data are coefficients followed by the absolute value of the t-statistic (in parentheses) below. All specifications include a constant term. In the two-stage estimates of regression 5, the stroke experience variable is instrumented by medical schools, business establishments, post offices, parks, and libraries. a Number is ×1012 for regressions 1–4. b Number is ×109 for regressions 1–4. *Statistically significant at the 10 percent level **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

dependent variable. Regression 1 presents the symmetrical specification: for the key independent variable, I use the number of sophisticated hospitals per capita. Possibly, however, the benefit to residents from these hospitals may not be a function of the number per capita. It might instead accrue simply from access to a sophisticated hospital—hence regression 2, with a dummy variable equal to 1 if an area has any sophisticated hospital. Or the benefit might increase with the number of such hospitals rather than the number per capita—hence regression 3, with a variable equal to the number of sophisticated hospitals. Because hospital experience will reflect in part the disease environment in the locality, these hospital experience variables may be endogenous. To explore this possibility, regression 4 uses a two-stage specification with the hospital experience variables instrumented through a

Chapter Four

120 Table 4.14 Hospital experience and mortality: Heart attacks Dependent variable: Heart attack deaths 1 OLS (per cap) Heart attack hospital Per capita Dummy

2 OLS (per cap)

Fraction 15–64 Fraction ≥65 Fraction >652 Densityb N Adjusted R 2

4 2SLS (per cap)

5 OLS (no.)

6 2SLS (no.)

−8.899** (2.32) −0.00006** (2.42)

Number Populationa

3 OLS (per cap)

−0.179 3.02 (0.77) (0.12) 0.00002 0.00007 (0.05) (0.14) 0.001* 0.001* (1.67) (1.69) 0.001 0.001 (0.62) (0.65) −6.58*** −6.80*** (2.99) (3.09) 855 855 0.34 0.34

−0.00005** −0.0001* (2.22) (1.65) 0.05 22.9 (0.00) (0.66) 0.00006 0.0001 (0.12) (0.22) 0.001* 0.001* (1.69) (1.66) 0.001 0.001 (0.63) (0.67) −6.81*** −6.99*** (3.09) (3.15) 855 855 0.34 0.33

−19.776*** −63.673*** (5.05) (5.29) 0.0003*** 0.0003*** (73.59) (51.96) −25.559 7.540 (0.32) (0.09) 635*** 618*** (4.27) (3.88) −1185*** −1130*** (4.18) (3.71) −0.002*** −0.002*** (5.72) (5.64) 855 855 0.88 0.86

Sources: Kosei rodo sho (2006a, 2006c); Somu sho (2007); Shobunsha (2007); Yomiuri (2008); Shukan Asahi (2008). Notes: Data are coefficients followed by the absolute value of the t-statistic (in parentheses) below. All specifications include a constant term. In the two-stage estimates of regression 5, the stroke experience variable is instrumented by medical schools, business establishments, post offices, parks, and libraries. a Number is ×1012 for regressions 1–4. b Number is ×109 for regressions 1–4. *Statistically significant at the 10 percent level **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

series of variables that capture a locality’s investments in public infrastructure more generally (see section D.4, above). Regressions 5 and 6 provide the simple, back- of-the- envelope estimates of the number of lives at stake. Because the number of deaths is on the left side of the equation and the number of sophisticated hospitals on the right, the coefficient on the hospitals gives the number of lives saved per hospital. Regression 5 reports a simple OLS model, and regression 6 reports the corresponding two-stage estimate. Note, however, that the use of a count measure as the dependent variable calls for the use of a negative binomial model rather than OLS. I use OLS instead merely because it simplifies the task of estimating lives saved.

Few Claims, but for a Different Reason

121

All these regressions include demographic controls. 55 I divide the population into three age groups (fi ner partitions are not available at the locality level) and omit the fraction aged less than 15. Although I include the fraction aged over 65, cities with relatively high fractions of over– age- 65 citizens might also have a relatively high fraction of the oldest citizens within this over- 65 group. To address this possibility, I include both the over-age- 65 fraction and its square. If the cities with the most over- 65 residents did also have larger fractions of the very oldest portion of the over- 65 group, the coefficient on this squared term should be positive. As tables 4.13 and 4.14 show, it is not. Mortality rates could also reflect the degree of crowding. To capture this possibility, I include the population density of an area. The calculated coefficient is consistently negative. (b) Strokes As table 4.13 shows, the coefficient on the stroke hospital variable is consistently and significantly negative in all specifications. Residents of areas with hospitals that use modern procedures are less likely to die of strokes. This is true regardless of whether I take the number of such hospitals per capita, a dummy equal to 1 if an area has any such hospital or the simple number of such hospitals. It is true regardless of whether I estimate the regression in OLS or instrument the stroke hospital variable in two-stage least squares. Regressions 5 and 6—with a dependent variable equal to the number rather than rate of stroke deaths—offer a simple way to gauge the magnitudes involved. According to the OLS estimate in regression 5, each sophisticated hospital cuts stroke mortality in the area by eight deaths per year. According to the two-stage estimate in regression 6, it cuts mortality by sixteen deaths. As the median city had fi fty stroke deaths a year (mean, 90), either estimate represents a substantial effect. (c) Heart Attacks As table 4.14 shows, hospitals with accumulated experience in angioplasties similarly reduce the risk of heart-attack deaths. Again, the effect is consistent and significant across all specifications. It holds regardless of whether I measure hospital sophistication by the number of experienced

122

Chapter Four

hospitals per capita, by a dummy variable equal to 1 for the presence of any such hospital, or by the simple number of such hospitals. And it holds in both OLS and two-stage estimates. Once more, the back- of-the- envelope estimates in regressions 5 and 6 suggest a substantial magnitude. According to the OLS estimate in regression 5, each hospital experienced in angioplasties reduces the number of fatal heart attacks by twenty. The two-stage estimate in regression 6 suggests each hospital saves sixty-four lives. As the median city had only twenty-nine fatal heart attacks (mean, 53), the estimates are unrealistic—but again do reflect the large size of the impact. F. Implications for Aggregate Costs Although the price caps directly reduce aggregate Japanese medical costs, these mortality patterns reduce it further still. The low caps obviously reduce the per-procedure cost of medical care. By eliminating many of the most severely ill patients, however, the mortality patterns reduce the total number of procedures billed as well. In effect, the low caps cut aggregate costs by accelerating the time that critically ill patients die. As many observers have noted, a substantial fraction of US health- care costs arises in the last few months of a person’s life (e.g., Cutler and Miara 1998, 2001). Typically, a patient undergoes a complex medical procedure at the outset of this last period. The procedure keeps the patient alive, who then over the next several months incurs additional costly expenses. Often, however, the deteriorating health that led to the fi rst procedure eventually kills the patient anyway. In Japan, that patient is less likely to have survived the initial procedure. Take cerebrovascular disease. After decades of saturated fats, lackadaisical exercise, and a stressful marriage, suppose a fi fty-year- old US executive suffers a stroke. The ambulance takes him to a large metropolitan hospital, and physicians operate on him at great cost. Through the operation, they save his life. Over the next several months, they continue the high- cost medical treatments that keep him alive. By the end of the year, though, the decades of bad food, sedentary living, and marital chaos causes a second stroke. This time, he dies. In part, medical costs are low in Japan because fewer such executives live to suffer the second stroke. More than in the United States, they die from the fi rst instead. Most communities in Japan do not even have so-

Few Claims, but for a Different Reason

123

phisticated emergency rooms, much less hospitals with extensive experience in cerebrovascular care. 56 When the Japanese executive has his fi rst stroke, his wife will call the ambulance. It will take him to the local emergency room, but most emergency rooms are a surgeon with an x-ray machine. The surgeon will give him a few drugs, and promptly he will die anyway.

V. Conclusions Universal health insurance boosts demand. Offer free medical care to people who would otherwise do without, and more people will bid for the services. All else equal, people will obtain a higher quantity of the service, and more resources will be spent on the service. Among wealthy capitalist societies, Japan spends relatively little on health care. In part, it caps its expenses by capping the amount it pays for each procedure at a quarter of what insurers pay in the United States. Some variation on such caps will probably inhere in any universal insurance program anywhere. Otherwise, voters would simply refuse to pay the bill. Yet, if caps inhere, so, too, will lower quality. Boost demand, limit supply, and cap prices—and sellers will respond by degrading quality. Some aspects of the quality degradation in Japanese medicine are well known: the long waits, the perfunctory consultations, and the need to return repeatedly for prescription renewals. Other effects are less well known. Because of the price suppression, the most talented science students will tend to opt out of medicine and go into fields like engineering. Within medicine, the most talented students will choose fields like cosmetic surgery that lie outside the national insurance system. Because the price suppression in Japan favors basic services over sophisticated modern procedures, fewer doctors will provide those sophisticated procedures. Fewer will learn how to do them at all. Fewer hospitals will offer them. Fewer patients with strokes and heart disease will receive them, and those who do receive them will obtain the procedures from doctors and hospitals with less experience. Chapter 5 turns to the implications this phenomenon poses for malpractice litigation. Medical malpractice is not randomly distributed across all medical procedures. Instead, disproportionately it accompa-

124

Chapter Four

nies the most complex, sophisticated, high-risk procedures. Because of the price suppression at the heart of the national insurance, Japanese doctors perform fewer such procedures. Necessarily, Japanese patients encounter less malpractice. Unlike traffic accidents and products liability, the low number of claims in the medical malpractice field does not result from the Japanese second-best approach to adjudication. It does not result from anything about the courts at all. Instead, as I explain in chapter 5, it results from the effect of the national health insurance system on the type of care that doctors supply.

Chapter Five

Medical Malpractice (II)

J

apanese fi le few medical malpractice claims. The reason does not lie in any characteristic of the legal system itself. It does not lie in any of the putative “problems” in the system claimed by its critics. Neither does it lie in any second-best approach discussed in chapters 2 and 3. Instead, the reason lies in the way the national health insurance system suppresses costs by limiting the amounts it pays doctors. In suppressing the amounts it pays for medical services, the Japanese national health insurance necessarily degrades quality. But it does not just degrade. It also skews the mix of services sold. Because it caps the prices it pays for technologically sophisticated services at particularly low levels, it alters the mix of procedures that doctors provide: more of the primitive medical care commanding the relatively higher prices and less of the technologically complex services commanding the relatively lower prices. Chapter 4 details two consequences of this pricing bias. For malpractice claiming patterns, the shift in service mix matters crucially. Patients seldom sue unless they experience an observable “bad outcome.” But observably bad outcomes do not as often occur when physicians see fundamentally healthy patients, work in small settings, and do relatively little of consequence. Neither do such doctors often commit legally defi ned malpractice. Instead, bad outcomes and provable malpractice occur when physicians see high-risk patients, work with a team of medical specialists, and undertake complex and technically sophisticated procedures. For obvious reasons, services that generate the most observable bad outcomes and provable negligence generate the most malpractice claims. If the national insurance causes sellers to supply fewer such services, patients will fi le fewer malpractice claims.1

126

Chapter Five

Chapter 5 explores this dynamic with aggregate data on Japanese malpractice suits and insurance premia, and microlevel data on all Japanese published medical malpractice opinions from 1995 to 2004. I fi rst describe the data and variables and then explore potential biases (section I). Through the data, I ask which doctors Japanese patients sue, what they collect, and why they choose these doctors to sue (section II). I estimate the amount of claiming and litigation (section III) and then discuss several unsatisfactory explanations for the low levels of claiming and suing (section IV.A). I conclude by explaining why the low levels instead reflect pathologies caused by the national insurance system (section IV.B).

I. Data A. The Databases To examine malpractice disputes in detail beyond that disclosed by the administrative office of the courts (and presented in chap. 4), I examine all judicial decisions published in the course of a decade. More specifically, I code every district court opinion published from 1995 to 2004 that appears in a search for “medical malpractice” in the Hanrei taikei database (also used in chaps. 2 and 3).2 This yields a population of 351 opinions, 348 civil and 3 criminal. With this information, I produce two datasets: a case-level database and a prefecture-level database. B. The Variables 1. case-level database I code each opinion for the following variables. Summary statistics appear in table 5.1. a. Financial Award value: Total amount awarded to the plaintiff. Demand value: Total amount demanded by the plaintiff.

Table 5.1 Selected summary statistics n A. Case level 1. Financial Award value (per 1000) Demand value (per 1000) 2. Delays Filing to judgment Accident to judgment 3. Recovery Plaintiff recovers No causation 4. Patient Male Death Age 5. Accident Misdiagnosis Medication error Surgery Obstetrics Emergency room Cardiac care Cerebrovascular Cancer 6. Institution University hospital Government hospital Red Cross hospital Other public hospital Private hospital Dental clinic Clinic B. Prefecture level 1. Explanatory variables Suits Population, per thousand Population aged >64 years, % Agricultural economy, % Density GDP per capita, per million Hospital beds Clinic beds Medical school Cardiac bypass Attorneys 2. Instruments Museums Concerts School Internet College grads

Min

Mean

Median

Max

31,100 74,400

14,500 59,400

205,000 546,000

4 5

11 14

1 0

1 1

1 1 38.5

1 1 88

346 338

0 589

265 265

0 1

348 343

0 0

.744 .201

341 348 316

0 0 0

.543 .592 36.3

348 348 348

0 0 0

.124 .152 .463

0 0 0

1 1 1

348 348 348 348

0 0 0 0

.098 .075 .103 .167

0 0 0 0

1 1 1 1

343 343 343 343 343 343 343

0 0 0 0 0 0 0

.175 .207 .047 .125 .227 .015 .187

0 0 0 0 0 0 0

1 1 1 1 1 1 1

46 46 46 46 46 46 46 46 46 46 46 46 46 46 46

Note: Published opinion database, as described in text.

0 607 16.4 .3 67.4 2.7 9,396 622 1 0 28 3.3 9.85 37 7.2

4.20 5.82

7.57 2,748 22.5 25.3 649 3.5 34,933 3,436 1.72 3.74 456 11.2 12.7 70.4 12.3

2.5 1,798 21.9 27.7 268 3.6 24,367 2,898 1 2 91.5 9.85 12.4 71.6 11.4

106 12,600 27.1 89.2 5751 6.5 129,939 10,990 13 20 10,263 32.9 15.6 92.1 24.2

128

Chapter Five

b. Delays Filing to judgment: Number of years from the year of fi ling to the judgment, provided fi led within three years of accident. Accident to judgment: Number of years from the year of accident to the judgment, provided fi led within three years of accident.

c. Recovery Plaintiff recovers: 1 if the plaintiff recovered at least some amount in damages; 0 otherwise. No causation: 1 if the court found that the defendant did not fully cause the accident, that the defendant did not fully cause the patient’s damages, or that the patient was partially negligent as well; 0 otherwise.

d. Patient Male: 1 if the patient is male; 0 otherwise. Death: 1 if the patient died from the accident; 0 otherwise. Age: Age of the patient at the time of the accident.

e. Accident Misdiagnosis: 1 if the wrongful act involved a misdiagnosis; 0 otherwise. Medication error: 1 if the wrongful act involved a medication error; 0 otherwise. Surgery: 1 if the wrongful act involved surgery; 0 otherwise. Obstetrics: 1 if the wrongful act involved obstetrics; 0 otherwise. Emergency room: 1 if the wrongful act took place in an emergency room; 0 otherwise. Cardiac care: 1 if the wrongful act involved cardiac care; 0 otherwise. Cerebrovascular: 1 if the wrongful act involved cerebrovascular disease; 0 otherwise. Cancer: 1 if the wrongful act involved cancer; 0 otherwise.

Medical Malpractice (II)

129

f. Institution University hospital: 1 if the wrongful act took place at a university hospital; 0 otherwise. Government hospital: 1 if the wrongful act took place at a government (but not university) hospital; 0 otherwise. Red Cross hospital: 1 if the wrongful act took place at a Red Cross hospital; 0 otherwise. Other public hospital: 1 if the wrongful act took place at any other public hospital; 0 otherwise. Private hospital: 1 if the wrongful act took place at a private hospital; 0 otherwise. Dental clinic: 1 if the wrongful act took place at a dental clinic; 0 otherwise. Clinic: 1 if the wrongful act took place at a clinic; 0 otherwise.

g. Other Year suit fi led, year of accident, and geographical dummies for the most often used district courts.

2. prefecture-level database At the prefecture level, I calculate the following variables: Suits: Number of malpractice suits fi led in the prefecture, 1995– 2004. Population: Population, 2005. Population aged over 64 years, %: Percentage of population age 65 or older. Agricultural economy, %: Value of agricultural output, divided by prefectural GDP Density: Population per square kilometer. GDP per capita: Prefectural GDP per capita, in billion yen. Hospital beds: Number of hospital beds (Kosei rodo sho 2008a). Clinic beds: Number of clinic beds (Kosei rodo sho 2008a).

Chapter Five

130

Medical school: Number of medical schools. Cardiac bypass: Number of hospitals performing more than one hundred heart surgeries (including cardiac bypass operations but not catheterization) in 2007 (Asahi shinbun shuppan 2009). Attorneys: Total number of attorneys, 2004 (Nihon bengoshi rengo kai 2005).

3. instruments As instruments for the number of attorneys per prefecture (see section VI.A.2.b), I add the following: Museums: Total museums in prefecture (including zoos, aquariums, etc.), 2002 (Toba 2005). Concerts: Percent of population (aged ≥10 years) who attend music concerts (for reasons not explained, the source excludes classical concerts), 2001 (Toba 2005). School Internet: Percentage of public schools with high-speed Internet access, 2003 (Toba 2005). College graduates: Percentage of population who graduated from a university, 2000 (Toba 2005).

C. Biases 1. introduction Like the American legal databases Lexis and Westlaw, the Hanrei taikei purports to include all published opinions. Some of these opinions appeared in one or more official (often subject-specific) court reporters. The rest appeared in the private commercial reporters. This database is biased. Whether a collection of all published malpractice opinions might be biased is not the question: clearly, it is. Publication introduces one obvious bias. Judicial administrative committees decide which opinions to publish officially. Staffed by career judges, they presumably publish those opinions that their members think provide (among other things) proper precedential direction. By contrast (and not to put too fi ne a spin on it), the publishers of the private reporters are in

Medical Malpractice (II)

131

the business of selling magazines. They include those opinions that they think will boost subscription rates. The very fact of litigation introduces a second bias. We have known at least since Priest and Klein (1984) that litigated cases—whether published or not—are not a random sample of all disputes. Instead, they represent those disputes that the parties chose not to settle out of court. Given that the vast majority of disputes settle, those that do not will potentially differ along several important dimensions. Consider, then, some evidence about the direction and magnitude of the biases involved. 2. published malpractice opinions represent a larger fraction of the underlying court cases than other published civil opinions From 1998 to 2004, the Japanese courts issued 2,298 civil judgments in medical malpractice cases (summing table 4.1, panel A, column 3). During the same period, the various reporters published 229 (10.0 percent; summing table 4.1, panel A, column 4). 3 By contrast, in 2004 court reporters published only 1.9 percent (1,358) of all civil judgments (as detailed in chap. 3). If I exclude default judgments, they published 3.0 percent (from Hanrei taikei database; Saiko saibansho, Shiho tokei 2004, table 20). Apparently, reporters publish about five times as many malpractice opinions as civil opinions more generally. A large fraction of the civil suits represent legally mundane disputes. As a relatively new field, medical malpractice opinions raise more interesting issues, and the reporters publish a bigger fraction of them (much the same phenomenon occurred with products liability, of course—see chap. 3). 3. plaintiffs in the published malpractice cases win more often than malpractice plaintiffs generally According to panel B of table 4.1, the fraction of medical malpractice cases that the published- opinion plaintiffs won rose from 65– 75 percent in 1995– 2001 to over 90 percent by 2003. During the same period, plaintiffs in malpractice cases as a whole—published and unpublished—won only 30–45 percent. Recall from chapter 4 that plaintiffs in US malpractice cases win about 20– 30 percent. Hypothetically, the high plaintiff win-rate in the published cases

Chapter Five

132

might reflect a desire among the elite judges in the administrative office to encourage malpractice claims. In fact, it does not. The official reporters nearly boycotted the malpractice field. During the ten years involved, they published only fi fteen opinions in civil malpractice opinions. Among them, the plaintiffs won only eight. The high plaintiff win-rates instead track the editorial bias at the commercial reporters. Perhaps the editors liked malpractice opinions but thought plaintiff losses bored their readers. Apparently, they decided that plaintiff victories offered subscribers a “better read” than plaintiff losses. Among the 337 cases in the private reporters, plaintiffs won 75.4 percent. 4. wrongful- death claims constitute a slightly higher percentage of the published malpractice cases than the unpublished Among the published cases, 59.5 percent involved patients who died (table 5.2). For all malpractice cases, the administrative office does not release the comparable fraction. Nonetheless, Hagihara, Nishi, and Nobutomo (2003) studied all malpractice decisions (N = 435) in ten district Table 5.2 Amounts recovered for published malpractice cases 3

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 1995–2004

4

1

2

Total Payouta

Death, %

Publisheda

Death (published)

979 834 564 1,041 957 1,181 1,408 1,338 1,485 954 10,761

69.0 68.2 72.7 57.9 69.6 47.9 52.3 50.0 46.7 62.5 59.5

23.3 19.4 17.1 27.4 41.6 25.3 32.0 44.6 49.5 59.6 31.1

38.9 51.4 36.5 55.8 61.9 53.6 63.0 38.3 49.7 74.7 50.0

5

Mean payout

Alla 24.0 15.3 15.4 40.1

Sources: Columns 1–4 are from the published opinion dataset, as described in the text. Column 5 is from Maeda, Sakamoto, and Nobutomo (2001) and gives the mean payout in all cases filed in court, including those that settled. Notes: Column 3 includes cases in which plaintiff did not recover. Column 4 is limited to wrongful death cases in which the plaintiff recovers at least some amount, and where the court does not reduce recovery for contributory negligence or causation issues. a Payout numbers are in million yen.

Medical Malpractice (II)

133

courts over 1986– 98. Examining both published and unpublished decisions, they found that 47.0 percent involved deaths. Apparently, the published opinions involve disproportionately many death claims.4 5. plaintiffs disproportionately sue in tokyo and osaka The administrative office of the courts does not disclose where malpractice plaintiffs bring their claims. In what seems the same procedure as Hagihara, Nishi, and Nobutomo (2003), however, Maeda, Sakamoto, and Nobutomo (2001) surveyed all (published and unpublished) malpractice cases from 1989–1998 for ten district courts (excluding Yokohama). They disclosed the number of cases from each court, and I report the numbers in table 4.1, panel C, column (3). Disproportionately, malpractice claimants sue in Tokyo and Osaka. According to the database used by Maeda, Sakamoto, and Nobutomo (2001), they fi led 41 percent of all malpractice cases in Tokyo and 27 percent in Osaka. Among civil cases more generally, plaintiffs fi led only 21 percent in Tokyo and 9 percent in Osaka (column 2). If anything, published malpractice opinions disproportionately include cases not from Tokyo or Osaka. Only 31 percent of the published cases come from Tokyo and 16 percent from Osaka (column 1). Recall, however, that Maeda, Sakamoto, and Nobutomo only surveyed ten district courts. Tokyo and Osaka cases are a larger fraction of their universe—but their universe includes fewer than all courts. For the most part, any geographical bias may not matter. Other than on time-to-judgment (see section I.C.6, below), the courts seem not to differ on any dimension measured here. 5 Indeed, Hagihara, Nishi, and Nobutomo (2003, 121) concluded that “there are no reports on regional differences in medical malpractice decision-making” among the different courts. The published- opinion dataset seems largely to confi rm this Hagihara, Nishi, and Nobutomo (2003) claim. Whether on the likelihood of recovery or the amount of damages received, the differences among the various courts are largely insignificant. 6. published malpractice cases take longer to adjudicate than the unpublished malpractice cases The plaintiffs in the published cases litigate longer than those in the unpublished cases. According to table 4.1, panel B, from 1998 to 2004,

Chapter Five

134

fi ling-to-judgment times for malpractice cases as a whole fell from about three years to two. During the same period, the fi ling-to-judgment times among the published cases decreased, but remained higher: from about 4.5 years to 3.5. The longer fi ling-to-judgment times in the published opinion database may reflect in part the slightly smaller fraction of Tokyo and Osaka cases (see section I.C.5, above). In table 5.3, column (3), I regress fi lingto-judgment time on the geographical variables. Parties that litigate their malpractice case in Tokyo can expect a decision nearly a year earlier than those who litigate it in one of the courts not listed. The longer fi ling-to-judgment times among the published cases may also reflect higher stakes. In my table 5.3 regression, I regress fi ling-tojudgment times on (inter alia) the amount the plaintiff demanded. The resulting coefficient is both positive and statistically significant. If the published cases involved higher stakes, they would (on average) take longer to try.

Table 5.3 Geographical differences

Dependent variable Tokyo Osaka Nagoya Yokohama Kobe Fukuoka Shizuoka Male Age Age2 Death No causation Plaintiff recovers Demand value n Adjusted/pseudo R 2 Regression

1 Plaintiff recovers

2 Award value

3 Filing to judgment

.025 (0.13) .479 (1.86)* .208 (0.54) −.078 (0.20) .288 (0.62) Dropped −.736 (1.49) .089 (0.53) .021 (1.84)* −.003 (1.97)* −.167 (0.96)

−5.95 (0.94) −9.55 (1.35) −12.1 (1.10) 3.96 (0.31) −.24 (0.02) −3.59 (0.33) −5.98 (0.26) 14.6 (2.91)*** .15 (0.41) −.007 (1.45) −6.09 (1.20) −30.1 (5.25)***

−.811 (2.56)** −.110 (0.31) 1.340 (2.54)** −.650 (1.18) −.060 (0.09) −.267 (0.42) 2.281 (2.35)** .179 (0.70) −.013 (0.76) −.000 (0.04) .432 (1.62) .574 (1.81) −.049 (0.16) 5.58a (2.42)**

299 .04 Probit

241 .18 OLS

239 .12 OLS

Notes: Coefficients in column 2 are divided by 106, and dataset is limited to plaintiff recovers = 1. Published opinion dataset, as described in text. a Number is ×109. *Statistically significant at the 10 percent level **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

Medical Malpractice (II)

135

7. published malpractice cases involve medical procedures similar to those among malpractice cases generally Though not comprehensively, the administrative office does disclose some information about the types of medical procedures that generate the malpractice suits. Of the 2005 suits, for example, 39 percent involved surgery. About fourteen percent involved obstetrics or gynecology (see table 4.2, panel A). The plaintiffs in the published opinions sued on a similar mix of suits. Of these opinions, the plaintiffs in 46 percent sued on surgical procedures. Fifteen percent sued on obstetrical or gynecological procedures. 8. addendum: comparison to settled cases (a) Yoshikawa In 2006, prominent malpractice plaintiff’s attorney Kozaburo Yoshikawa published a book detailing forty-five malpractice cases he had settled (Yoshikawa and Makabe 2006). Over the course of his career (1978– 2005), Yoshikawa explained, he had fielded about 510 inquiries from clients or potential clients. Of those, he had pursued about 100. The rest he had concluded showed too low a probability of success. Of those more than 100 cases, he then detailed the major cases that resulted in a plaintiff recovery: forty-five cases that settled and twelve cases that went to a court decision. Given the obvious incentive facing Yoshikawa to exaggerate his career success, these cases are decidedly nonrandom. That said, he reported: •

Of the 45 cases, 26 (58 percent) involved death claims (published opinion

•

database: 59 percent). Fifty-three percent of the claimants were men, with a mean age of 42 years (published opinion database: 54 percent men, with a mean age of 36 years).

•

The plaintiffs in the wrongful- death cases received settlements that ranged from 2 million to 80 million yen, with a mean of 29.7 million yen and a median of 30 million. Of the six cases with sub–10-million-yen recoveries, three were cancer cases (hence, plaintiffs with a low life expectancy) and two involved patients over age 70 (published opinion database: recoveries of 200,000 to 189 million yen, with a mean of 40.6 million yen and a median of 37.5).6

•

The nineteen claimants with nondeath claims received settlements that ranged from 5 million to 70 million yen, with a mean of 40 million and me-

Chapter Five

136

dian of 50 million (published opinion database: 200,000 to 205 million yen, with a mean of 43.6 million and a median of 20.7 million).

(b) Tokyo District Court Other evidence does indeed suggest that Yoshikawa’s settled cases probably include unrepresentatively generous settlements. The Tokyo District Court examined all cases in its malpractice panel that settled during April 2001 to September 2002 (Tokyo chiho saibansho 2003, 35– 36, 45). In table 5.4, I compare these settlements with the comparable statistics for the published opinion database. Note two facts about these settlements. First, the plaintiffs settled for a smaller fraction of the amounts they demanded than the litigating plaintiffs eventually obtained. Of the settling plaintiffs, 36 percent obtained at least half their demand. Of the published opinion plaintiffs, over 51 percent did. This is, of course, exactly what one would expect if parties settled for amounts that reflected a discount for their likelihood of recovery. Second, the settling plaintiffs obtain relatively small amounts. Only 27 percent of the settling plaintiffs recovered 20 million yen. Of the published opinion database, a full 58 percent did. This, too, is what one would expect if settling plaintiffs discounted their expected recovery by the likelihood of success—and is consistent with the general phenomenon of smaller-stake disputes settling more readily than larger-stake disputes.

Table 5.4 Settlement and litigation, by award/demand Award/demand, %

Settled

Litigated