Litigation and Social Capital: Divorces and Traffic Accidents in Japan

• DOI: http://doi.org/10.1111/jels.12034
• Author: J. Mark Ramseyer
• Date: 01 March 2014
• Published date: 01 March 2014

Litigation and Social Capital: Divorces and

Traffic Accidents in Japan

J. Mark Ramseyer*

Using regression and factor analysis with prefecture-level data, I ask whether Japanese in

communities with high levels of “social capital” more readily settle their disputes out of

court. Although studies of litigation rates often measure suits per capita, the more appro-

priate measure may involve suits per “dispute.” We lack information about the number of

disputes in many fields, but we do have it for Japanese divorces and traffic accidents—and I

focus on those two sets of disputes. Disputes over divorce and traffic accidents differ funda-

mentally, and social capital does not lower litigation rates among either. I find that:

(1) couples in communities with low social capital are more apt to divorce; (2) couples in

low-social-capital communities are not more likely to litigate their disputes; (3) couples in

communities with more lawyers are not more likely to litigate their divorces; and (4) parties

in communities with low social capital are not more likely to litigate their disputes over traffic

accidents; but (5) parties in communities with more lawyers are indeed more likely to litigate

their disputes over those accidents.

Litigiousness “is born of a breakdown in community,” Jethro Lieberman (1983:187) once

wrote. We sue because we no longer trust each other to pay voluntarily. We attack because

we no longer think anyone will help us when we need it. Once upon a time, we cooperated

first and reneged only when we found ourselves exploited. No longer. Rather than play the

venerable tit-for-tat, we defect from the start. We sue because, to use Robert Putnam’s

(2000) phrase, we have destroyed our stock of “social capital.” We defect because we “bowl

alone.”

Or so we tell ourselves. And in contrast to this (perhaps mythically) anomic and

litigious America, we have long presented Japan as a relatively cooperative world. Japanese

do indeed seem more cooperative than Americans, on any of several measures. Yet even

within Japan itself, those in some communities more often litigate than others. Might this

internal variation in the way Japanese use courts reflect variations in the level of Putnam’s

social capital?

Plausibly, social capital could affect the use of the courts in two very different ways.

First, communities with high levels of social capital might have fewer disputes. People might

*Harvard Law School, Cambridge, MA 02138; email: ramseyer@law.harvard.edu. The author is Mitsubishi Professor

of Japanese Legal Studies, Harvard University.

The author gratefully acknowledges the helpful suggestions of the editors, three anonymous referees, Wered

Ben-Sade, Tom Ginsburg, Glenn Hoetker, and workshop participants at the University of Lousanne, Bar Ilan

University, Tel Aviv University, and the generous financial assistance of the Harvard Law School.

bs_bs_banner

Journal of Empirical Legal Studies

Volume 11, Issue 1, 39–73, March 2014

39

start fewer brawls at the bar. They might steal less property. They might keep more of their

promises. If social capital is high, people might wrong each other less. Suffering fewer

wrongs, they might file fewer suits.

Second, given a fixed stock of disputes, people in communities with higher levels of

social capital might take a smaller fraction of those disputes to court. They might talk with

each other more willingly. They might compromise more quickly. They might reach

agreement more readily. They might cut enforceable deals more easily.

In the article that follows, I use the diversity within Japan to explore this possibly

double role that social capital might play. I use regression and factor analysis to ask whether

social capital correlates with the number of divorces. With the number of divorces held

constant, I ask whether social capital correlates with the tendency to negotiate a divorce

agreement out of court. I find that couples in communities with high levels of social capital

do divorce less frequently. Given a stock of divorces, however, those in communities with

high social capital are more—not less—likely to litigate their divorce. The availability of

attorneys has no effect on whether they litigate.

I contrast these results with litigation over serious traffic accidents. In traffic disputes

as well, social capital seems not to reduce litigation rates. The availability of an attorney,

however, does matter: the more accessible an attorney, the more likely the parties will take

a dispute to court.

Too often, in discussing “litigation rates” we conflate the question of (1) what causes

the disputes we observe with the question of (2) given those disputes, what causes people

to litigate them? Typically, we define the litigation rate as a variation on (litigated suits)/

(population). Yet that fraction itself combines two very different components: (litigated

suits)/(total disputes), and (total disputes)/(population).1

These two component fractions raise conceptually distinct issues. Where one variable

might increase the number of disputes in a community, another might increase the fraction

of those disputes that people choose to litigate. In this article, I use the prefectural variation

in social capital to explore (very preliminarily) the two questions. Using data on divorces

and serious traffic accidents, I ask whether social capital correlates with the frequency of a

dispute and with the fraction of those disputes that people choose to litigate in court.

The results are suggestive—but only suggestive. Obviously (but crucially), the results

from Japan need not generalize to other societies. Even within Japan, more finely parti-

tioned data would have generated more reliable results. Unfortunately, for most of the

relevant variables the data are available only at the prefectural level. Information on change

over time would also have increased reliability—but unfortunately again, many of the

variables are measured only occasionally at best. With obvious caveats about these limita-

tions, however, I exploit the variation among Japanese prefectures, and use regression and

factor analysis to explore the role that social capital may (or may not) play in the generation

and settlement of disputes.

1The work was pioneered by Felstiner et al. (1981), but the resulting literature is now quite large (e.g., Jacob 1992).

Not only does the level of injury potentially depend on the level of social capital, the question of whether people

decide to treat a given injury as a dispute could also—hypothetically—depend on that social capital.

40 Ramseyer

I begin by surveying the literature on litigation and social capital (Section I). I ask

whether the level of social capital correlates with divorce rates (Section II). I ask whether

it correlates with settlement rates in divorce disputes (Section III), and I conclude by

asking whether it correlates with settlement rates in disputes over serious traffic accidents

(Section IV).

I. Litigation and Social Capital

A. The Classic Model

Begin with the standard—now classic—model of litigation and settlement (Landes 1971;

Posner 1973). Stripped of refinements, it posits that plaintiffs can profitably settle their

claims for any amount (Pmin) greater than the estimated value they place on their claim

(Vp), less their expected litigation costs (Lp) but plus their expected settlement costs (Sp):

PVLS

pppmin =−+

Analogously, defendants can profitably pay in settlement any amount (Dmax) less than the

estimated value they place on the plaintiff’s claim (Vd), plus their own expected litigation

costs (Ld) and less their expected settlement costs (Sd):

DVLS

dddmax =+−

The parties face a range of mutually advantageous settlements when Pmin <Dmax. They

have no choice but to litigate when Pmin >Dmax. The more optimistic they each might be, the

more likely Pmin >Dmax and the more likely they will litigate. The greater the expected costs

of litigation, the more likely they will settle, and the lower the cost of settlement, the more

likely they will settle.

Consider now the effect of education and wealth. Educated disputants will find it

easier than others to manipulate complex bureaucracies. They will find it easier to under-

stand elaborate legal rules. In effect, they face lower litigation costs, and should litigate a

larger fraction of their disputes.

Richer parties should also litigate more of their disputes. By the terms of the classic

model, the parties can settle if but only if:

VLS VLS

pppddd

−+

Rearrange the terms, and

VV LL SS

pd pd pd

−< +

()

−+

()

Vpand Vdeach represent a sum of estimated probabilities (the odds that the plaintiff will

win) times possible judgments (the amounts the plaintiff might collect if he does win). Hold

constant the estimated probabilities but increase the range of possible judgments, and

Vp−Vdwill rise.

Divorces and Traffic Accidents in Japan 41

Increase the amount at stake, in other words, and the probability of litigation will rise.

Although disputants do rationally invest more resources in high-stakes disputes than low,

they do not usually increase litigation costs dollar-for-dollar with the amount at stake. As a

result, they will more likely face a viable settlement window in a small-stakes dispute than in

a large. Given that rich disputants will have more high-stakes disputes than poor, rich

disputants will have more disputes that they cannot cost effectively settle.2

B. Social Capital

1. Introduction

According to several prominent scholars, where social capital is high people trust each

other to cooperate and keep their word. They rely on communal norms to enforce any

agreements they negotiate. When the stock of capital falls, they lose that trust. Unable

anymore to enforce their agreements through social ties, they turn to law, lawyers,

and courts. Where social capital is high, they cooperate. Where social capital is low,

they sue.

2. Social Capital

Lieberman (1983:186) put the postulated tie between social capital and litigation most

starkly.

Litigiousness is not a legal but a social phenomenon. It is born of a breakdown in community, a

breakdown that exacerbates and is exacerbated by the growth of law. . . . [U]ntil there is a

consensus on fundamental principles, the trust that is essential to a self-ordering community

cannot be.

Social capital falls; litigation climbs.

Putnam explained the connection as a function of the trust that comes from a dense

network of social ties and obviates the need for the more costly legal apparatus. “[F]ormal

contracts, courts, litigation, adjudication, and enforcement by the state” are, he writes

(2000: 144–45), “one alternative to generalized reciprocity and socially embedded

honesty.” Since 1970, that expectation of generalized reciprocity has fallen in the

United States, and “informal understandings no longer [seem] adequate or prudent”

(2000:147).

Instead, Putnam (2000:21) argued, modern Americans manipulate legal processes to

their detriment.

A society characterized by generalized reciprocity is more efficient than a distrustful society, for

the same reason that money is more efficient than barter. If we don’t have to balance every

exchange instantly, we can get a lot more accomplished. . . . When economic and political dealing

is embedded in dense networks of social interaction, incentives for opportunism and malfeasance

are reduced.

2As found, for example, in Eisenberg et al. (2012) and Ginsburg and Hoetker (2006).

42 Ramseyer

As Elinor Ostrom (1998:16) put it, groups can “fail to achieve mutually productive benefits

due to their lack of trust in one another or to the lack of arenas for low-cost communication,

institutional innovation, and the creation of monitoring and sanctioning rules.”

Robert Ellickson (1991) made much the same claim in his study of Shasta County

farmers. “[T]he more close knit a group,” argued Ellickson (1991:250), the more successful

it will be at generating and enforcing utilitarian norms to govern internal disputes.”

Because members will consider their norms superior to the law, close-knit groups will tend

(1991:250) “to have controller-selecting norms that discourage members from taking

intermember disputes into the legal system.” Concluded Ellickson (1991:251):

Groups with large or transitory membership are usually not close-knit and cannot rely as much on

informal social control. As a result, resort to the legal system tends to be tolerated more in

industrialized than in preindustrial cultures, and more in large cities than in small towns.

In his controversial Coming Apart, Charles Murray (2012) argues that social capital is

not disappearing in the United States across the board. Instead, it remains intact in

professional communities, but has almost entirely vanished among the working class. He

calls the former “Belmont,” after the affluent Boston suburb, and the latter “Fishtown,”

after the working-class Philadelphia neighborhood of that name. In America’s Belmonts,

social capital remains high, writes Murray—and people read the newspaper, vote, attend

churches and synagogues, volunteer at the schools, marry, bear children within marriage,

and stay married; in Fishtown, social capital has vanished. With social capital high, in

Belmont people keep their promises; with the social glue gone, in Fishtown they rely on

partners at their peril. In Belmont, people trust; in Fishtown, they cheat.

3. Law and Social Norms

Legal scholars discussed several of these themes in the “law and social norms” literature.

Ellickson (1998:546) himself noted at the time that Putnam’s social capital was “analogous

to a set of adaptive norms.” He and others observed several points relevant here. As

Ellickson (1991:546, 1990) put it, “informal systems of control” are “especially” important

“where interacting parties have a continuing relationship” at stake. If a population is less

stable, norms are less secure.

Second, people follow norms to retain the chance to trade. Communities enforce

their norms, wrote Richard Posner, if they can tie them to an “implicit threat of ostracism,

that is, of refusal of advantageous transactions.”3If people engage in fewer transactions with

each other, norms are less secure.

Third, gossip matters. “Norms and rules, whether publicly or privately created,

embody and convey information,” explained Avery Katz.4

3(1997:366); see also Posner and Rasmusen (1999), Posner (1998:554), and Akerlof (1976:1980).

4(1996:1749); see Akerlof (1976) and Ellickson (1991:180–81).

Divorces and Traffic Accidents in Japan 43

They cannot be followed unless information is transmitted regarding their substantive content;

they cannot be enforced unless information is transmitted regarding who has obeyed them, who

has violated them, and who is to impose any associated punishment or reward.

If people do not know what their neighbors do, norms are less secure.

C. The Empirical Challenge

1. Introduction

The question at stake is whether to incorporate this social capital into the empirical inquiry.

The concept appears nowhere in the classic economic model of litigation and settlement—

for the model was never about trust or generalized reciprocity. It was about how people

behaved if the relationship disintegrated.

When Robert Mnookin and Lewis Kornhauser (1979) first applied the economic

model to divorce, for example, they did not try to introduce social capital. Instead, they

focused on the way court-enforceable rules shape the agreements couples reach out of

court. They focused, as they inimitably put it, on the way that couples “bargain in the

shadow of the law.”

To be sure, one could incorporate the concept of social capital into the economic

model—if one wanted. The question is whether one should want. The concept has been

central to segments of the law and society field, and one could readily turn it central to the

economic model as well.

Take a few of the simplest variations. Communities with high levels of social capital

maintain wide-ranging and swift channels of communication. Necessarily, they have lower

Spand Sd. Communities with high levels of social capital can cheaply punish members who

renege on their agreements. By freeing people from some of their worries about enforce-

ability, they again lower Spand Sd. Some tightly bound communities (think Amish) main-

tain norms against litigating disputes. If they punish those who sue, the expected

punishment could enter the model as an element subtracted from Pmin and added to Dmax

(or as a higher Lpand Ld).

As with the debate over “law and social norms,” the crucial question is empirical:

How much do factors like these matter observationally? That we live in a socially con-

structed world has been a commonplace among freshman sociology courses for decades

now. We take our norms and expectations—our “preferences,” in law and economics—

from the communities within which we live and work. We know that. We could add those

norms and expectations to the model, but is it worth it? In the end, the question is

whether the model’s increased explanatory power justifies the analytic complications it

inevitably introduces.

2. Wealth and Social Capital

To explore the empirical ties between social capital and litigation rates, we will need

necessarily to address several sets of colinearities. For example, social capital correlates with

wealth—and wealth affects litigation rates independently (see Section I.A). Perhaps in some

44 Ramseyer

societies the two factors correlate negatively. One could imagine a world where the poor live

in stable, closed rural villages, and the rich in anomic urban enclaves. One could imagine

a society where poor communities enjoy high levels of social capital, and the rich have

none.

One could imagine such a world, but in the modern United States we do not inhabit

it. Neither does anyone in Japan (see Section II.D.6). As Putnam (2000:193) put it,

“[p]eople with lower incomes and those who feel financially strapped are much less

engaged in all forms of social and community life than those who are better off.” As Murray

(2012) explores at elaborate length, the U.S. professional class lives within close networks

that provide high levels of information, where members make massive investments in

personal reputations, where they keep their families largely intact, and where they tightly

enforce norms of appropriate behavior. Those outside the professional class live with less

of this.

3. Education and Social Capital

Social capital correlates with education too. Education also correlates with wealth, of

course, and with litigation rates besides. Educated workers earn high incomes, and earning

high incomes accumulate greater wealth. In part, they earn higher incomes because those

incomes include a return on the investments they made in skills (think physics, calculus,

and the ability to write a coherent paragraph) valued highly on the market, and in part they

earn higher incomes because their level of education reflects a cognitive ability itself valued

highly on the market. They were able to finish their education, after all, because they

brought an intellectual capacity that employers value independently of anything they

learned in school.

Education also correlates with social capital independent of its correlation with

wealth. Education, wrote Putnam (2000:108), is “the strongest predictor of formal commu-

nity involvement.” Indeed (2000:118):

[H]ighly educated people . . . are more likely to volunteer, to donate money, and to give blood.

In particular, education is one of the most powerful predictors of virtually all forms of altruistic

behavior, even after controlling for other possible predictors.

Again, Murray (2012) makes much the same observation. His social-capital-

rich Belmont is not just wealthier than the social-capital-deprived Fishtown. It is also

much more highly educated. Indeed, he sorts communities in part by income, but in

part by education as well. His claim that Americans segregate themselves by the cogni-

tive ability reflected in their education embodies the very claim that made his book so

controversial.

4. Social Capital Itself

Social capital is not itself a measurable variable. Even those scholars determined to include

it in their studies find it hard to capture empirically. They have yet to settle on a best proxy

Divorces and Traffic Accidents in Japan 45

(or set of proxies) for it. Given the inherent vagueness of the concept, it is hard to see how

they ever could.

Because of the nebulous character of social capital, I follow Putnam and take as broad

an approach as I can (Putnam 2000:414, original in italic):

No single source of data is flawless, but the more numerous and diverse the sources, the less likely

that they could all be influenced by the same flaw. Two independent (though necessarily imper-

fect) strands of evidence are better than one, and more than two are better still, especially if they

have different imperfections.

Consistent with this maxim, I choose 13 variables that (loosely to be sure) reflect

different aspects of social capital: civic engagement, social engagement, workplace engage-

ment, community infrastructure, family cohesion, and community cohesion. I explain the

data and define the variables more precisely in Section II.D.

5. Data from Japan

My data from Japan cover two broad sets of disputes. As noted earlier, the level of social

capital in a community could affect both (1) the number of disputes and (2) the fraction

of those disputes that community members choose to litigate. Although we usually know the

number of disputes people litigate, we often have no idea how many underlying disputes

there might have been. About divorces and traffic accidents, however, we do know the

number of disputes: the number of couples who file for divorce and the number of serious

accidents.

In Japan, the city hall counts the total number of divorces; for this study, that figure

represents the number of disputes, and the ratio of divorces to marriages represents the

frequency of disputes. The courthouse counts the number of divorces that couples litigate.

The ratio of litigated divorces to total divorces then gives us the frequency of litigated

resolutions. With this information, we can estimate the effect of social capital both on

(1) the fraction of marriages that end in divorce and on (2) the fraction of divorces that

couples litigate in court.

Similarly, police statistics detail the number of traffic accidents per year; for this

study, that figure represents the number of disputes. Through the mid-1990s (but no

longer), court statistics disclosed the number of suits that the families of accident victims

filed (Saiko 1994:tabs. 10, 23). Combine the two numbers, and we can estimate the effect

of social capital on the fraction of major traffic disputes that people litigate in court.

Although extremely common, divorces and traffic accidents epitomize different

classes of disputes. They differ, moreover, in ways potentially tied to the role that social

capital might play. The parties to a divorce know each other extremely well; the parties to

a traffic accident seldom know each other at all. Divorces involve the central unit in a

community—the family; accidents often involve people from far distant cities and villages.

Divorces can concern children; accidents often do not. All these factors could affect

how social capital shapes—or does not shape—the way people resolve their various

disputes.

46 Ramseyer

II. Divorce in Japan

A. Introduction

Begin, then, with the relationship between social capital and the decision to divorce. In the

analysis below (Section III), I take the decision to divorce as given, and ask whether the level

of social capital correlates with a tendency to negotiate that divorce privately. Preliminarily,

however, consider whether that level of social capital correlates with the decision about

whether to divorce at all: Are couples in some communities more likely to divorce than

couples elsewhere?

Start with the Beckerian model of marriage and divorce (Section II.B), and the

practical mechanics of divorce in Japan (Section II.C). Turn to the data on social capital

and divorce (Section II.D). Finally, consider the empirical exercise: the correlation between

that social capital and divorce (Section II.E).

B. The Classic Model

Consider Gary Becker’s classic discussion of marriage and divorce (1991:Ch. 10). Becker

posits that couples will stay married if, but only if, they enjoy a higher level of combined

welfare together than apart (1991:331). If they enjoy greater total welfare within marriage,

they will remain married. If they would enjoy greater welfare outside of marriage, they will

divorce.

Whether two people earn higher aggregate welfare in marriage will turn in part on

the investments they make in the relationship (1991:329). Suppose they invest heavily in

skills, qualities, and assets (think children) that pay higher returns if their marriage

endures. They will more likely find their combined welfare higher in marriage, and avoid

divorce. In turn, they will make higher levels of those marriage-specific investments if they

each think their partner will not explore his or her “outside” options.

Social capital can provide the “glue” necessary to prevent a spouse from exploring

those alternative arrangements. If he (or she, the logic runs in both directions) probably

will not demand a divorce, she can safely invest in marriage-specific skills and assets. If she

makes those marriage-specific investments, they will both earn returns from their marriage

that they would lose if they divorced. And if they earn returns in marriage that they would

forfeit under divorce, they will less likely divorce.

Consider how social capital might give that “glue.” In communities with high levels

of social capital, couples may have large networks of mutual friends that would unravel

in a divorce. They may find fulfillment in a religious institution that discourages (or

perhaps bans) divorce. If the institution also bans remarriage, it again slashes their divorced

welfare.

More broadly, a couple in a community with high social capital may live within deep

and broad-ranging networks of friendships, family ties, and other social relations. If these

networks discourage divorce (as many networks in high-social-capital communities do),

each spouse will be less likely to explore his (or her) divorced alternatives. If one spouse

does not explore those alternatives, then the other can more safely invest in the marriage,

Divorces and Traffic Accidents in Japan 47

and if they each invest in marriage-specific skills and assets, they will together generate

the returns necessary to keep their combined welfare higher within marriage than

without.5

C. The Practical Mechanics

Most Japanese couples negotiate their divorces outside the courts (Bryant 1984). To marry

in Japan, they enter their tie in the “family registry” (koseki) at their (usually local) city (or

ward, village) hall. To divorce, they simply enter the event in the same registry. Provided

they reach an agreement that satisfies both, they can divorce with no more government

involvement than a sheet of paper before a petty city hall bureaucrat.

Only couples who cannot agree on the terms of their divorce go to court. Once there,

the judge will usually route them to a mediator. If they reach an agreement before the

mediator, the judge will order the divorce on those negotiated terms. If they cannot agree,

the judge will try the divorce claim in court.

On a population of 126 million in 2010, Japanese concluded 700,000 marriages and

251,000 divorces. Of these divorces, the couples negotiated 220,000 out of court (kyogi).

They reached 25,000 agreements through court mediation (chotei), and another 3,600

during court proceedings (wakai). The courts adjudicated 2,500 divorce disputes through

trial. The ratio of divorces to marriages ranged from 28.9 percent in Tokyo to 45.2

percent in Aomori prefecture. The fraction of divorces negotiated as private agreements

ranged from 80.4 percent in Yamagata prefecture to 92.1 percent in Okinawa (Kosei

2010a).

Note that for U.S. divorces, we have the number of disputes but (in most states) not

the number of claims litigated. Like Japanese couples, most U.S. couples negotiate their

divorce privately. As Mnookin and Kornhauser put it (1979:951), “the overwhelming major-

ity of divorcing couples resolve distributional questions concerning marital property,

alimony, child support, and custody without bringing any contested issue to court for

adjudication.” But even when they negotiate the deal on their own, most U.S. couples still

file their divorce in court for what Mnookin and Kornhauser call the judicial “rubber

stamp.” U.S. sources give the total number of divorces. Generally (with some exceptions),

they do not distinguish between those divorce decrees where the judge resolved significant

questions, and those where he simply pressed the “rubber stamp.”

D. The Data

1. Introduction

To explore the impact of social capital, I begin with a dependent variable that captures the

ratio of divorces to marriages. I add three sets of independent variables. For reasons

5As Becker (1991:329) notes: “Expectations about divorce are partly self-fulfilling because a higher expected prob-

ability of divorce reduces investments in specific capital and thereby raises the actual probability.”

48 Ramseyer

outlined earlier (Section I.A), the classic model of litigation and settlement suggests that

litigation rates should rise with both wealth and education. Accordingly, I identify several

proxies for the wealth and educational levels of a community. I then turn to the inquiry at

the heart of this study: proxies for the amount of social capital. I include summary statistics

in Table 1.

Note several qualifications. First, I have information only at the prefectural level

(these are the Japanese equivalent of the U.S. states). Although I would prefer a more finely

grained data set, I obtain my data from a wide variety of sources, and for most of the

variables the information is not available below the prefectural level. In turn, the limited

number of observations necessarily restricts statistical “power” and the reliability of the

results.

Table 1: Selected Summary Statistics

Min Median Mean Max

Dependent Variables

Divorce ratio 0.289 0.374 0.375 0.452

Divorce settlement rate 0.804 0.871 0.869 0.921

Suits per death 0.186 0.461 0.585 2.424

Economic Variables

Income PC (×10,000) 325 436 437 600

Savings PC (×1,000) 5,068 15,647 14,887 19,577

High income TP (×1,000) 0.149 0.330 0.406 1.881

Education Variables

Advance to high school 92.8 96.7 96.5 98.4

Advance to university 30.1 43.7 43.1 53.5

Expenses PC elem. education 761 905 924 1,211

Expenses PC high school 941 1,141 1,171 2,173

Social Capital Variables

Voter turnout 38.2 58.8 59.5 69.4

Newspaper readership 0.22 9.52 11.3 25.4

Volunteering 19.7 27.6 28.2 34.5

Religious followers PC 0.625 1.485 1.720 6.473

Job tenure 8.7 11.6 11.5 13.0

Unemployment 3.2 4.7 4.8 7.5

Com’ty staff PC (×1,000) 0.02 0.11 0.13 0.44

Crime rate 4.99 10.01 10.50 18.87

Illegitimacy rates 1.29 2.10 2.15 3.99

Abortions PC (×1,000) 0.293 2.07 2.04 2.89

Population change −0.087 −0.027 −0.018 0.083

Unemployment coll. grads. 7.5 21.5 21.8 42.7

Metropolitan 0 0 0.191 1

Other

Attorneys PC (×1,000) 0.058 0.094 0.126 1.076

Scriveners PC (×1,000) 0.093 0.154 0.154 0.270

Notes: Thetable gives summary statistics for the key variables. See text for precise definitions of the variables and the

sources used.

Divorces and Traffic Accidents in Japan 49

Second, several of the sources do not measure the relevant variables every year.6

Indeed, the “high-income taxpayer” database (see Section II.D.3) is no longer available at

all. This obviously prevents me from taking a dynamic approach that explores change over

time. It also means, however, that for some of the variables I do not have data from exactly

the same year. For the most part, however, the relative wealth, education, and social capital

of Japanese prefectures is stable. I know of no reason to think that the relative ranking

among prefectures of wealth, education, or social capital would have changed in any

substantial way over the last decade.

Although the 13 proxies for social capital cover a wide range of phenomena, they

correlate with each other closely. Of the 72 possible pairs among them, the correlation is

statistically significant at the 10 percent level in 49. Apparently, most of the variables

capture the same basic pattern of interprefectural variation. I include correlation coeffi-

cients in Ramseyer (2012:tab. 2).

Given the empirical challenge to having 13 highly correlated proxies for an unmeas-

ured latent variable (particularly with data available only at the prefectural level), I use

factor analysis to create a set of synthetic measures of social capital. I take the 13 correlated

proxies for the latent variable, and reduce them to two factors (following standard practice,

I keep only those factors with eigenvalues greater than 1). I create similar synthetic meas-

ures for economic and educational levels. I then regress the several dependent variables on

the synthetic measures.

2. Dependent Variable

In the regressions below, I look for correlates of the decision by married couples to divorce.

I define that decision by:

•Divorce ratio: The number of divorces in 2010, divided by the number of

marriages in 2010. I take the data from the Japanese government’s vital statistics

database (Kosei 2010a).

3. Independent Variables—Economic

I define three proxies for the level of economic welfare in a prefecture:

•Income PC: Mean annual salary in 2008 in 10,000 yen. I take the data from Nenshu

(2012).

•Savings PC: Mean bank deposits (but not other wealth) per capita in 2004, in

1,000 yen. I take the data from Somu sho (2004).

6This is not the case for all the variables, of course. Many of the most basic statistical measures are compiled every year

by the Japanese government. Increasingly, the data are available online. This is true for the Health Ministry’s massive

“vital statistics” database (Kosei 2010a) (downloadable as CSV files), for example, but also for the Education Ministry’s

survey of educational expenses (Monbu 2008), and the Police Agency’s white papers (Keisatsu 1994). The court’s

litigation statistics (Saiko 1994) are also available online, but only as a set of coarsely partitioned summary statistics on

pdf files. For other material, availability over the Internet varies. The Scriveners’ Association publishes its data on its

website (Nihon shiho 2012), for example. By contrast, the Bar Association makes available its white papers only for

purchase (Nihon bengoshi 2011).

50 Ramseyer

•High-income taxpayers PC: The number of high-income taxpayers per capita in

2004. The National Tax Office (which released the data through 2004) defined a

high-income taxpayer as anyone who paid 10 million yen or more in income tax.

To pay that much tax in 2004, a taxpayer would have needed to earn about

$400,000. If only the very wealthiest couples litigated their divorces, then this

variable would explain litigation rates more fully than the other two economic

variables. I take the data from Tokyo (2004).7

4. Independent Variables—Education8

I include four proxies for the level of education:

•Advance to high school: The percentage of middle school (Grades 7–9) stu-

dents in 2002 who advanced to high school (Grades 10–12). Education in Japan is

compulsory only through Grade 9. I take the data from Toba (2005:89).

•Advance to university: The percentage of high school students in 2002 who

advanced to university. I take the data from Toba (2005:90).

•Expenses per student,elementary school: The mean amount spent on each

elementary school student in 2008, in 1,000 yen. I take the data from Monbu

(2008).

•Expenses per student,high school: The mean amount spent on each high

school student in 2008, in 1,000 yen. I take the data from Monbu (2008).

5. Independent Variables—Social Capital

a. Civic Engagement. As proxies for social capital, I begin with voter turnout rates and

newspaper subscriptions. The former represents the most straightforward measure of civic

engagement: How many people participate in elections? The latter measures how hard

people try to learn what they need to know to vote intelligently. Putnam (2000:36), for

example, calls newspaper readership a measure of popular “interest in politics and current

events.”

•Voter turnout: Voter turnout rates for the national election in 2003. I take the

data from Somu sho (2003).

•Newspaper readership: The percent of households subscribing to the morning

edition of the Asahi newspaper from July to December 2011. The Asahi is the

closest thing Japan has to a “newspaper of record.” With a diffusion rate of 11

percent of the households, it undersells the Yomiuri (15 percent) but the latter

lacks the Asahi’s “gravitas.” The diffusion rate of the Mainichi is 6 percent. The

7For a fuller discussion of the high-income taxpayer data set (which, after 2004, is no longer available), see Nakazato

et al. (2009, 2010, 2011) and Ramseyer (2009).

8For a discussion of the relative merits of controlling or not controlling for education in measuring changes in social

capital, see Putnam (2000:418–19).

Divorces and Traffic Accidents in Japan 51

correlation coefficients among the three papers range from 0.62 to 0.80; all are

significant at better than 0.1 percent. I take the data from Nihon ABC (2011).

Unfortunately, the newspaper data omit observations for the three northeastern

prefectures of Iwate, Miyagi, and Fukushima. This causes the number of observations in the

regressions involving this variable to fall from 47 to 44 (see, e.g., Table 5).

b. Social Engagement. To capture the extent to which people participate in their commu-

nity, I examine volunteering rates and religious affiliation. The former captures people’s

willingness to help keep their community intact. The latter reflects a subsidiary facet of the

same willingness. According to Putnam (2000:67), religious participation functions “as a

powerful correlate of most forms of civic engagement”; according to Murray (2012:207), it

constitutes “one of the key sources of social capital in a community.”

•Volunteering: Percentage of residents 15 years or older performing volunteer

work, 2006. I take the data from Somu sho (2006).

•Religious followers PC: The number of members of a religious organization

(of any faith) in 2008, per capita. I take the data from Bunka cho (2008).

c. Workplace Engagement. “Many people form rewarding friendships at work,” writes Putnam

(2000:87–88), and “feel a sense of community among co-workers, and enjoy norms of

mutual help and reciprocity on the job.” As a result, those with low job tenure enjoy less

“trust and social connectedness in the workplace.” Those who lose work completely with-

draw psychologically from the community itself. The financial stress from their unemploy-

ment places “a profoundly depressing effect on social involvement” (Putnam,

2000:192–93), and leads to “less time spent with friends, . . . less frequent attendance at

church, less volunteering, and less interest in politics.”

•Job tenure: Average number of years on a job, 2009. I take the data from Gekkyu

(2009).

•Unemployment rate: Unemployment rate in 2010. I take the data from

Somu sho (2012).

d. Community Infrastructure. Through their government, citizens can invest in the facilities

and staff necessary to keep their community cohesive. They can hire people to run com-

munity centers, and they can hire police officers to keep order. About the police, Putnam

(2000:308) writes:

Higher levels of social capital, all else being equal, translate into lower levels of crime. . . . States

with more social capital have proportionately fewer murders. . . . This inverse relationship is

astonishingly strong—as close to perfect as one might find between any two social phenomena.

I examine staffing patterns at community centers, and rates of serious crime.

•Community center staff PC: The number of staff at local community centers

(kominkan) in 2005, per capita. I take the data from Monbu (2005).

52 Ramseyer

•Crime rate: The number of violations of the Criminal Code (excluding traffic

violations) in 2010, per 1,000 population. I take the data from Homu sho (2011).

e. Family Cohesion. For measures of family cohesion, I turn to illegitimacy and abortion

rates. Bronislaw Malinowski (1962:137) called his “principle of legitimacy” a universal

sociological law. All societies insist, he controversially explained, that:

no child should be brought into the world without a man—and one man at that—assuming the

role of sociological father, that is, guardian and protector, the male link between the child and the

rest of the community.

Within the modern United States, children born out of marriages (born “illegiti-

mate”) do not fare well. As Murray (2012:164) put it, their mothers “come disproportion-

ately from the lower socioeconomic classes and they tend to provide worse environments

for raising children than married mothers.” What is more, if mothers are bearing their

children outside of marriage, then families are necessarily playing a smaller role in the

community—yet “families with children are the core” of well-functioning communities

(Murray, 2012:165). If illegitimacy is high, social capital is low.

Abortions represent a costly way to limit family size. For the most part, men and

women in cohesive families use other ways to prevent pregnancies. If abortions are up,

social capital may be down (though I recognize the possibly controversial nature of this

claim).

•Illegitimacy rates: Percentage of children born outside of marriage, 2010.

I calculate the rate from data available at Kosei (2010a).

•Abortions PC: Abortions in 2007, per capita. I calculate the rate from data

available in Kosei (2008).

f. Community Cohesion. As measures of community cohesion, I take the growth rate of a

community and the job prospects of its educated youth. Cohesive communities grow. Their

residents thrive. They attract others, and retain young people who might otherwise emi-

grate. A community’s growth rate will thus reflect its relative attractiveness. What is more,

because educated men and women contribute disproportionately to a community’s stock of

social capital, the job prospects for the best-educated youth matter heavily for a comm-

unity’s ability to cohere as an integrated unit. Should those job prospects fall, educated men

and women will simply leave.

•Population change: Fractional increase in population, 2000 to 2010. I take the

population numbers from Kosei (2010a).

•Unemployment,college graduates: Unemployment rate among new college

graduates, 2002. I take the data from Toba (2005:75).

g. Other. Putnam (2000:206) suggests that “[l]iving in a major metropolitan agglomeration

somehow weakens civil engagement and social capital.” Hence:

Divorces and Traffic Accidents in Japan 53

•Metropolitan: 1 if the prefecture has one of the nine most populous cities in

Japan in 2012; 0 otherwise.

6. Interregional Variation

On these measures, Japanese prefectures show substantial variation. As homogeneous as

Japan may seem on the surface, the country incorporates major contrasts—in wealth, in

education, and in social capital. Japan was once an agricultural society with a few large

cities. It is no longer. Over the past century (and particularly over the past half-century)

people (and particularly the most talented people) have migrated in massive numbers to

the urban centers. There, they have reached advanced educational levels, earned large

incomes, and built communities with all the markers of high levels of social capital. They

have left behind rural villages of men and women who are often poor, badly educated, and

without the social glue that once tied them together.

Few pairs of prefectures exhibit as massive a contrast as Okinawa and Tokyo. Okinawa

epitomizes the periphery; Tokyo represents the center. Okinawa lies at the southern tip of

Japan; Tokyo sits at the heart of the country. Okinawa was ruled by others; Tokyo ruled

others. Okinawa fell under the power of a Japanese clan in the 17th century, and became

a prefecture in the 19th; Tokyo has held the seat of government since 1600.

Okinawa is poor; Tokyo is rich. Okinawans earn the lowest income per household in

Japan: 3.24 million yen. Tokyo residents earn the highest—6.00 million—and families in

Osaka, Aichi (i.e., the city of Nagoya), and Kanagawa (i.e., Yokohama) earn incomes close

to those in Tokyo. Okinawan households hold the smallest bank accounts in Japan—5.07

million yen—and families in the southern island of Kyushu hold similarly small accounts

(Miyazaki, Kagoshima, Nagasaki). Tokyo families control the biggest bank accounts: 19.6

million.

Okinawa has no jobs; Tokyo recruits. At 7.5 percent, unemployment in Okinawa is

the highest in the country. In Tokyo, unemployment sits at 5.5 percent. If Okinawans do

find a job, they do not keep it as long as other Japanese. At 8.7 years, the mean job tenure

in Okinawa is the shortest in the country (followed closely by the Kyushu prefectures of

Miyazaki, Kagoshima, and Fukuoka). In Tokyo, it is 11.1 years.

Okinawans drop out of school; Tokyo offers the best universities in the country. Of all

high school graduates, those from Okinawa are least likely to go to college: barely 30.1

percent attend. Of those from Tokyo, 52.4 percent do—similar to those in the urban

centers of Hiroshima (51.9 percent), Hyogo (i.e., Kobe, 52.5 percent), and Kyoto (53.5

percent). When they do finish college, Okinawan students find it extraordinarily hard to

find jobs. Among its recent graduates, 28.7 percent were unemployed (similar to the

Kyushu prefectures of Fukuoka, Oita, and Kumamoto). Among Tokyo graduates, only 21.5

percent were.

World War II brutalized Okinawa. The U.S. army and marines landed on the islands

in April 1945, and fought for two months. About 100,000 Japanese soldiers died, and

somewhere between 40,000 and 150,000 Okinawa civilians died too. Apparently, some

Japanese army commanders may have shot civilians lest they spy for the Americans. Some

may have ordered civilians to commit mass suicide rather than surrender to the Americans.

54 Ramseyer

And Okinawans have never let the rest of Japan forget. The scope of the travesties

remains a dispute (Hata 2009). National bureaucrats and politicians routinely cite low

estimates of wartime civilian deaths. Okinawan politicians and commentators insist on high,

and use the claims (together with the ostensible continuing insult from the U.S. bases) to

demand large subsidies. The transfer payments from the national government are indeed

enormous (Hook 2003:42–44). Six decades after the end of the war and four after the

reversion to Japan, the prefecture remains hugely dependent on the national fisc.

The poverty and low schooling in Okinawa suggest low social capital; the wealth and

education in Tokyo suggest high. Consistent with this, Okinawa has the third lowest number

of religious followers per capita in Japan (the Kyushu prefecture of Miyazaki is fourth).

Metropolitan (and apparently secular) Tokyo has the third highest (Hyogo, with Kobe city,

is fifth). Okinawa has the highest rate of shotgun marriages: 42.4 percent (the Kyushu

prefectures of Saga and Kumamoto also make the top five). Tokyo has the lowest: 19.5

percent (the metropolitan prefectures of Kyoto, Aichi, and Kanagawa round out the top

four).9Okinawa has the highest illegitimacy rate in the country: 3.99 percent. At 1.97

percent, the rate in Tokyo is half that.

Although the data show substantial interregional variation, the detail suggests an

obvious reason any conclusions may not generalize to other societies: even in seemingly

anomic Okinawa, the apparent levels of social capital seem much higher than in other

wealthy democracies. Illegitimacy rates in Okinawa may be double those in Tokyo, but they

still lie under 4 percent. In the United States, more than 40 percent of all births are to

unmarried mothers—54.8 percent in Mississippi, 65.5 percent in Puerto Rico, and 71.6

percent in the Virgin Islands (Centers for Disease Control and Prevention 2012). Forty-two

percent of marriages in Okinawa may be “shotgun,” but the couples do marry. In most

sectors of the United States, the very term carries the oddly anachronistic flavor of Dwight

Eisenhower and Fred MacMurray television shows.

E. Results

1. Introduction

Later in this article I focus on whether levels of social capital predict out-of-court settle-

ment; I ask whether—given divorce—couples negotiate or litigate their divorce decrees.

Preliminarily, however, consider whether social capital predicts the divorce itself. It does.

Whether according to regressions directly on the various proxies for social capital, or

according to a regression on a synthetic measure for social capital constructed through

factor analysis, higher levels of social capital predict lower divorce ratios.

These preliminary results suggest that the proxies reflect real aspects of the under-

lying social capital. If the proxies did not predict divorce rates, that failure would suggest

that perhaps they did not reflect the latent variable. In Sections III and IV, I find that they

9Shotgun marriages are percent of marriages (earlier of cohabitation or filing of marriage on family registry) that

occur less than nine months before the birth of a child, 2009. See Kosei (2010b).

Divorces and Traffic Accidents in Japan 55

do not predict settlement rates. If they did not predict divorce ratios either, I would worry

that perhaps they did not correlate with social capital. In fact, they predict divorce rates

quite nicely.

2. Individual Proxy Variables

In Tables 2 and 3, I regress the divorce ratio on each of the independent variables discussed

above. The regression is ordinary least squares (I include generalized linear model esti-

mates in Table 5), and the data are prefecture level. For the divorce ratio, I take the number

of divorces in a prefecture in 2010, and divide it by the number of marriages (Kosei 2010a).

Much as the model predicts, several of the variables correlate with a lower divorce

ratio. First, wealthier communities enjoy more stable marriages (Table 2, Panel A). The

calculated coefficients on wealth, income, and high-income taxpayers are all negative at

statistically significant levels. Becker (1991:336–37) and Murray (2012:156) find the same

phenomenon in the United States. Given that richer communities have higher levels of

social capital, the regressions here may capture that omitted variable.

Second, education has a more ambiguous effect. Most of the calculated coefficients

in Panel B of Table 2 are insignificant, though higher rates of college attendance do

correlate with lower divorce rates. The coefficients on the other three variables are positive

(albeit insignificant). Arguably, advancement rates to high school and college capture

different social phenomena: the fraction that attends college measures the modern upper-

Table 2: Economic and Educational Correlates of Divorce

Dependent Variable Divorce Rate

A. Economic

Savings PC −6.40***

(3.93)

Income PC −0.327***

(3.76)

High-income TP −63.468***

(3.18)

Adj R20.24 0.22 0.17

B. Educational

Adv to high sch 0.003

(0.58)

Adv to university −0.003***

(4.06)

Exp per stud elem educ 0.854

(1.67)

Exp per stud high school 0.040

(0.13)

Adj R2−0.01 0.25 0.04 −0.02

***p<0.01; **p<0.05; *p<0.10.

Notes: The table gives the result of an OLS regression of divorce ratios on economic and educational levels.

Coefficients, followed by the absolute value of the tstatistic. All regressions include a constant term. Coefficients on

Savings are ×1,000,000, on Income are ×1,000, on Exp per stud elem education and Exp per stud high school are

×10,000. n=47. See text for definitions and sources of variables.

56 Ramseyer

Table 3: Social Capital Correlates of Divorce

Dependent Variable Divorce Rate

Civic Engagement

Voter turnout −0.0009

(0.84)

Newspaper rdp −0.0008

(0.96)

Social Engagement

Volunteering −0.002

(1.24)

Relig follwr PC −0.010**

(2.13)

Workplace Engagement

Job tenure −0.015**

(2.02)

Unemployment 0.019***

(3.16)

Community Infrastructure

Comm’ty ctr PC −43.273

(065)

Crime rate −0.002

(1.21)

Family Cohesion

Illegitimacy 0.044***

(6.23)

Abortions PC 22.557*

(1.90)

Community Cohesion

Populatn growth −0.527***

(4.20)

Coll unemploy’t 0.002**

(2.60)

Other

Metropolitan −0.017

(1.25)

Adj R2−0.01 −0.00 0.01 0.07 0.06 0.16 −0.01 0.01 0.45 0.05 0.27 0.11 0.01

***p<0.01; **p<0.05; *p<0.10.

Notes: The table gives the result of an OLS regression of divorce rates on measures of social capital. Coefficients, followed by the absolute value of the tstatistic. All regressions include

a constant term. See text for definitions and sources of variables.

Divorces and Traffic Accidents in Japan 57

middle professional class, whereas the fraction that fails to attend high school measures the

dysfunctional lower class. Yet where advancement rates to high school are positively corre-

lated with the per-student expenditures, advancement rates to university are negatively

correlated. Where advancement rates to university are positively correlated with incomes,

savings, and the presence of high-income taxpayers, high school rates are negatively cor-

related. I do not have an explanation for these results.

Third, consistent with theory, divorce rates are inversely correlated with several

indices of social capital (see Table 3). For example, divorce rates are higher in communities

with fewer religious followers; shorter job tenure; higher unemployment rates; higher

illegitimacy rates (see also Figure 1); more abortions; lower rates of population growth; and

higher unemployment rates among college graduates. Divorce rates are low, in other words,

where adults—especially educated adults—can find jobs, where both parents focus on

raising their children, and where families find the environment attractive enough to choose

to settle.

3. Factor Analysis

As an alternative to regressing the divorce ratio directly on these proxies for social capital,

I use the proxies to estimate the underlying level of social capital itself. As in all studies on

this topic—including the well-known Putnam (2000) and Murray (2012)—social capital

constitutes an unmeasured latent variable. Factor analysis offers a way to combine its various

proxies into a synthetic variable that more closely approximates it. As Stata (2005:213)

explains, the procedure “reduces the number of variables in an analysis by describing linear

combinations of the variables that contain most of the information and . . . admit mean-

ingful interpretations.” I use the standard “principal factors” estimation procedure, with

Figure 1: Divorce rates and illegitimacy rate.

0.25

0.3

0.35

0.4

0.45

1234

Divorce Rate

Illegitimacy

Divorce Rate

Divorce Rate

Notes: Thefigure shows the positive association between a community’s divorce ratio and illegitimacy rate. The prefecture

with the highest illegitimacy rate is Okinawa; the prefecture with the lowest divorceratio is Tokyo. See text for sources.

58 Ramseyer

“orthogonal varimax” rotation. Following general practice, I retain those factors with

eigenvalues greater than 1. Table 4 reports the rotations.

I begin by using the three proxies for economic welfare to estimate a common

economic factor. As Panel A of Table 4 shows, the factor loads all three proxies heavily. I

then estimate a common educational factor (Panel B). Consistent with the puzzle described

above (Section II.E.2), the factor combines the two educational advancement rates in

opposite directions: high school advancement positively but university advancement nega-

tively. Troublingly, both advancement variables have high uniqueness values. Uniqueness,

explains Stata, measures “the percentage of variance for the variable that is not explained

by the common factors.” When the figure is high, it suggests that “the variable is not well

explained by the others” (Stata 2005:215).

Table 4: Rotated Factor Loadings

A. Economic Factor

Variable Econ Fr Uniqueness

Income PC 0.8950 0.1990

Savings PC 0.6935 0.5191

High-inc TP 0.7189 0.4832

B. Educational Factor

Variable Educ Fr Uniqueness

Advance to HS 0.5244 0.7250

Advance to univ −0.3158 0.9003

Exp elem ed 0.7054 0.5024

Exp high sch 0.4434 0.8034

C. Social Capital Factors

Variable Factor1 Factor2 Uniqueness

Voter turnout 0.6371 −0.4187 0.4188

Newspaper read 0.1570 0.6270 0.5822

Volunteering 0.5249 −0.5789 0.3893

Rel fllowers .4141 −0.1156 .8151

Job tenure 0.7824 0.0123 0.3877

Unemployment −0.8255 0.2773 0.2416

Com’ty staff 0.4540 −0.4744 0.5689

Crime rate −0.1084 0.8215 0.3134

Illegitimacy −0.8736 −0.1103 0.2247

Abortions −0.1333 −0.4864 0.7457

Population ch −0.1383 0.7260 0.4538

Unemp coll grad −0.7077 0.1226 0.4841

Notes: The table gives the rotated factor loadings for the synthetic

measures of economic welfare, educational levels, and levels of social

capital. Principal factors method used. Orthogonal varimax rotation

results reported. Factors retained if eigenvalue >1. See text for defini-

tions and sources of variables.

Divorces and Traffic Accidents in Japan 59

In Panel C of Table 4, I use the 12 continuous social capital proxies to produce two

factor variables. The first heavily loads the work-related variables to produce a variable that

rises with levels of social capital: it combines job tenure positively, and general and college-

graduate unemployment negatively. Consistent with this approach, it loads illegitimacy

rates negatively as well. Factor 2 borders on theoretical incoherence: that it loads crime

rates positively and voter turnout and volunteering negatively suggests that it falls with social

capital; that it loads population change and newspaper circulation positively suggests it rises

with social capital.

In the first column of Table 5, I regress the divorce ratio on the synthetic measures

and on metropolitan (OLS in Regresssion 1; generalized linear model with logit link,

binomial family in Regression 1(a)). The result is straightforward: divorce falls with social

capital. Consistent with theory (Section II.B), the coefficient on the first social capital factor

is negative and strongly significant: as social capital rises, divorce ratios fall. The educational

and second social capital factors reflect loadings that make little theoretical sense, and the

regression produces statistically insignificant coefficients for both.10

III. Divorce Litigation in Japan

A. Introduction

Theory predicts that divorce ratios should correlate inversely with levels of social capital,

and so my data show. The proxies for social capital in modern Japan do indeed predict

divorce ratios. They predict it whether I use the proxies directly, or use synthetic measures

created from those proxies through factor analysis.

Do the same proxies predict whether divorcing couples litigate their divorce

agreement?

B. Dependent Variable

I capture the decision to settle a divorce out of court with the following variable:

•Settlement rate: The number of divorces negotiated out of court (kyogi),

divided by the total number of divorces. I take both figures from Kosei (2010a).

I include summary statistics in Table 1. Elsewhere, I use other definitions of a negotiated

divorce settlement, and find that they generate the same results (Ramseyer 2012:tab. 6).

C. Results

1. Individual Proxy Variables

a. Wealth. By the classic model of litigation and settlement, wealthier communities

(because their disputes involve higher stakes) should have higher litigation rates. Other

10Although power analysis (powerreg in Stata) indicates that the sample size is more than adequate to test the effect

of the first social capital factor (with a power of more than 0.95), the same size is too small to test the effect of the

second factor.

60 Ramseyer

Table 5: Regression Results on Synthetic Factors

Dependent

Variable

(1) Divorce

Ratio

(1a) Divorce

Ratio

(2) Divorce

Settlement Rate

(2a) Divorce

Settlement Rate

(3) Divorce

Settlement Rate

(4) Suits

per Death

Econ factor −0.008 −0.033 0.011** 0.087** 0.007 −0.248

(0.85) (0.57) (2.67) (2.51) (0.88) (1.68)

Educ’l factor 0.008 0.034 −0.004 −0.032 −0.006 −0.060

(1.05) (0.97) (1.25) (1.53) (1.33) (0.74)

SC factor 1 −0.022*** −0.093*** −0.018*** −0.163*** −0.016*** 0.104

(3.85) (2.63) (7.21) (6.72) (4.17) (1.30)

SC factor 2 −0.002 −0.007 −0.006 −0.047 −0.005 0.084

(0.17) (0.16) (1.43) (1.58) (1.10) (0.95)

Metropolitan −0.009 −0.039 −0.003 −0.028 −0.003 0.260*

(0.60) (0.62) (0.43) (0.56) (0.50) (1.90)

Attorneys PC 21.848 5446.732***

(0.62) (4.56)

Adj R20.42 0.59 0.55 0.66

OLS GLM OLS GLM 2SLS 2SLS

***p<0.01; **p<0.05; *p<0.10.

Notes: Thetable gives the results of regressions of divorce rates, divorce settlement rates, and traffic accident litigation rates on measures of economic welfare, educational

levels, and levels of social capital. Regressions 1 and 2 are OLS; 1a and 2a are a generalized linear model with a logit link, binomial family; 3 and 4 are two-stage least squares

with Attorneys PC instrumented by Scriveners PC. Coefficients, followed by the absolute value of the tand zstatistics. All regressions include a constant term. n=44. The

numbers for attorneys and scriveners (the instrument) are for 2010 and 2012 in the third column but for 1997 in the last.

Divorces and Traffic Accidents in Japan 61

empirical studies—whether of Japan (Ginsburg & Hoetker 2006) or elsewhere (e.g.,

Eisenberg et al. 2012)—do present some results consistent with this logic. In Panel A of

Table 6, I regress the settlement rate on three indices of economic welfare. The results are

uniformly insignificant (note, however, the significant results in the factor analysis; Section

III.C.2).

b. Education. By the classic model, couples in communities with high educational

levels should also (because they have lower effective litigation costs) have higher litiga-

tion rates. In Panel B of Table 6, I regress the settlement rate on prefecture-level

educational variables: the coefficient on the rate at which middle-school students

advance to high school is significantly and negatively associated with settlement rates.

The larger the fraction of students who attend high school, the more people litigate

their disputes (see Figure 2 for a graphical illustration). Similarly, the more heavily a

prefecture invests in elementary school education, the more likely its couples will litigate

their divorce (note, however, the insignificant results in the factor analysis; Section

III.C.2).

Table 6: Economic and Educational Correlates of

Divorce Settlements

Dependent Variable Settlement Rate

A. Economic

Savings PC 0.946

(0.95)

Income PC 0.41

(0.78)

High-income TPs PC 17.656

(1.54)

Adj R2−0.01 −0.00 0.03

B. Educational

Adv to high sch −91.64***

(3.79)

Adv to university −0.529

(0.11)

Exp per stud elem educ −0.498

(1.84)*

Exp per stud high school 0.041

(0.25)

Adj R20.23 −0.02 0.05 −0.02

***p<0.01; **p<0.05; *p<0.10.

Notes: The table gives the result of an OLS regression of divorce

settlement rates on economic welfare and levels of education. Coeffi-

cients, followed by the absolute value of the tstatistic. All regressions

include a constant term. Coefficients on Savings are ×1,000,000, and on

Adv to high sch, Adv to univ, Exp per stud elem educ, and Exp per stud

high school are ×10,000. n=47. See text for definitions and sources of

variables.

62 Ramseyer

c. Social Capital.

i. Introduction. Lieberman (1983) and Putnam (2000) suggest that people in communities

with high levels of social capital will more likely settle their disputes than people in

low-social-capital communities. If so, then a regression of the settlement rate on proxies for

social capital should yield significantly positive coefficients. It does not.

Instead, the data from Japanese divorces suggest that Lieberman and Putnam have

the relationship exactly backward. When statistically significant, the relationship between

the proxies for social capital and the settlement rate is reversed: the higher the level of

social capital in a community, the lower the settlement rate. Analogously, the higher the

social capital, the higher the rate of litigation.

ii. Civic engagement. I begin Table 7 with measures of civic engagement. I turn first to voter

turnout rates—the most direct measure of that civic engagement. Exactly counter to the

Lieberman-Putnam logic, turnout rates are negatively associated with settlement rates: the

more actively citizens engage in the democratic process, the more likely they will litigate

rather than settle their divorces out of court.

Newspaper readership is not significantly associated with litigation rates.

iii. Social engagement. With variables proxying for social engagement, the association

between social capital and settlement rates again runs opposite to the Lieberman-Putnam

conjecture. In communities where citizens volunteer for local responsibilities, they sue (i.e.,

do not settle). Where they participate in local temples and shrines (religious followers

PC), they sue.

Figure 2: Divorce settlement rate and high school advancement.

0.8

0.85

0.9

0.95

92.5 93 93.5 94 94.5 95 95.5 96 96.5 97 97.5 98

Settlement Rate

Advance HS

Settlement Ra te

Settle ment Rate

Notes: Thefigure shows the negative association between a community’s divorce settlement rate and educational level. See

text for sources.

Divorces and Traffic Accidents in Japan 63

Table 7: Social Capital Correlates of Divorce Settlements

Dependent Variable Settlement Rate

Civic Engagement

Voter turnout 0.002***

(4.26)

Newspaper rdp −0.0004

(0.08)

Social Engagement

Volunteering −0.002***

(3.50)

Relig follwr PC −0.005*

(1.83)

Workplace Engagement

Job tenure −0.017***

(5.13)

Unemployment 0.010***

(3.17)

Community Infrastructure

Comm’ty ctr PC −99.179***

(3.08)

Crime rate 0.002**

(2.55)

Family Cohesion

Illegitimacy 0.023***

(5.97)

Abortions PC −0.713

(0.11)

Community Cohesion

Populatn growth 0.157**

(2.17)

Coll unemploy’t 0.001***

(3.11)

Other

Metropolitan 0.015**

(2.07)

Adj R20.27 −0.02 0.20 0.05 0.36 0.16 0.16 0.11 0.43 −0.02 0.07 0.16 0.07

***p<0.01; **p<0.05; *p<0.10.

Notes: The table gives the result of an OLS regression of divorce settlement rates on levels of social capital. Coefficients, followed by the absolute value of the tstatistic. All regressions

include a constant term. Coefficients on Savings are ×1,000,000. See text for definitions and sources of variables.

64 Ramseyer

iv. Workplace engagement. With workplace engagement, too, higher levels of social capital

are associated with lower settlement rates (higher litigation rates). In prefectures with long

job tenure, the settlement rate is low and—by implication—the litigation rate high. In

prefectures with low unemployment, the settlement rate is low—and the litigation rate

high.

v. Community infrastructure. The relationship between the level of community infrastruc-

ture and the settlement rate tracks the results in the preceding panels. Where people invest

heavily in staffing their community centers (their kominkan), couples litigate their divorces.

Where police keep crime levels low, couples litigate their divorces.

vi. Family cohesion. I also explore measures of family cohesion. Whether couples use abor-

tions to control pregnancies is not correlated to settlement rates. The extent to which they

bear children out of marriage is: the greater the illegitimacy rate (the lower the social

capital), the higher the settlement rate.

The relationship between illegitimacy and settlement is extremely strong (see

Figure 3 for a graphical illustration). Scholars such as Murray (2012) stress the extent to

which illegitimacy captures social dysfunction. According to Table 7, that dysfunction

correlates positively not with litigation, but with out-of-court settlement.

vii. Community cohesion. Table 7 shows two other ties between proxies for social capital and

the settlement rate. First, unemployment rates among college graduates are positively

associated with settlement rates. The harder graduates find it to locate a job, the more

people settle divorces out of court.

Figure 3: Settlement rate and illegitimacy rate.

0.8

0.85

0.9

0.95

012345

Settlement Rate

Illegit imacy

Settlement Ra te

Settle ment Rate

Notes: Thefigure shows the positive association between a community’s divorce settlement rate and illegitimacyrate. See text

for sources.

Divorces and Traffic Accidents in Japan 65

Second, the more a community has grown (the more it thrives), the more people

settle their divorces. Here (and only here), the association between social capital and

settlement is positive—consistent with the Lieberman-Putnam conjecture.

viii. Urban character. Finally, Putnam suggests that urban communities enjoy less social

capital than do rural ones. According to Table 7, couples in the most urban prefectures

(those with the largest cities) are most likely to settle their disputes.

2. Factor Analysis

With respect to social capital, factor analysis confirms these results. In Column 2 of Table 5

(OLS; generalized linear model in Regresssion (2a)), I regress the divorce settlement rate

on the synthetic measures created through that analysis. Note that the coefficient on the

economic factor is positive and significant: couples in rich communities are more likely to

settle their divorce. As noted earlier, the educational factor borders on theoretical inco-

herence (Section II.E.2), and generates no statistically significant effect.

The coefficient on the first social capital factor is statistically very significant, and in

the opposite direction from that suggested by Lieberman (1983) and Putnam (2000).

Recall that the factor loads heavily on voter turnout, job tenure, illegitimacy, and the two

unemployment variables. According to the Column 2 (Table 5) regression, the higher this

first social capital factor, the lower the settlement rates. Like the educational factor, the

second social capital factor is theoretically incoherent (Section II.E.3), and generates no

significant effect.11

3. Conclusion

These results present a puzzle. I do not know why higher levels of social capital should be

associated with lower levels of out-of-court settlement—and, by implication, with higher

levels of litigation. Perhaps the proxies for wealth and education fail to capture the full

extent of the variation along those dimensions (as the discussion above suggests; Sections

II.E.2, III.C.2). If so, then the various measures of social capital could incorporate some of

the effect of wealth and education. Alternatively, perhaps the divorces in high-social-capital

communities are more bitter than those where social capital is low. After all, couples divorce

more readily where social capital is low—perhaps many were triggered by a lower level of

animosity. If couples only rarely divorce where social capital is high, perhaps those divorces

that one observes involve higher emotional and economic stakes.

The more basic point, however, is clear: the data show no sign that social capital

increases out-of-court settlement. At least in the context of modern Japanese divorces,

prefecture-level data show no evidence of any such effect.

11Although power analysis (powerreg in Stata) indicates that the sample size is more than adequate to test the effect

of the first social capital factor (with a power of more than 0.95), the same size is too small to test the effect of the

second factor.

66 Ramseyer

D. The Presence of Attorneys

1. Introduction

Potentially, the extent to which couples litigate their divorces might turn on their access to

professionals who could handle their claims. If so, then any attempt to explain litigation

rates should account for the availability of attorneys. Because their number is endogenous

to the use of litigation, however, ordinary least squares is inappropriate.

In their pioneering work on Japanese prefectural litigation rates, Tom Ginsburg and

Glenn Hoetker (2006) instrument the number of attorneys by the population of a distinct

group of nonlawyer legal professionals (judicial scriveners) and the high court in whose

jurisdiction the prefecture falls. They explain:

We use the high court district for each prefecture (a set of seven dummy variables) and the

number of shiho shoshi (legal Scriveners) in the prefecture as identifying variables. While attorneys

and judges may have preferences regarding the high court district in which they want to practice,

we expect that few potential litigants will be aware of what high court district they are in, much less

be aware of the implications of those districts for litigation they might pursue. Shiho shoshi should

not affect the amount of litigation, as litigation is limited to bengoshi [attorneys]. However, since

the services of shiho shoshi and bengoshi overlap somewhat (for example, legal advice), the presence

of shiho shoshi would influence the economic attractiveness of a prefecture to an attorney.

In this article, I take a similar approach: I instrument the number of attorneys per capita by

the number of judicial scriveners per capita. Readers should bear in mind, however, that

instrumental variables regressions do not have good small-sample properties, and the N

here is only 47.

2. Variables

I add the following additional variables.

•Attorneys PC: The number of attorneys per capita, 2010. I take the data from

Nihon bengoshi (2011:84–85).

•Scriveners PC: The number of judicial scriveners per capita, 2012. I take the data

from Nihon shiho (2012).

3. Results

I present the results of the two-stage estimates in Table 8: the number of lawyers has no

effect on the litigation rate. Instead, the coefficient on attorneys PC is consistently

insignificant. As in the earlier results, the coefficient on savings PC is significantly negative

in several specifications: the richer the prefecture, the more litigation (i.e., the less settle-

ment). The coefficient on advance to high school is significantly negative in all seven

specifications: the better educated the prefecture, the more litigation. Voter turnout and

volunteering are both significantly negatively associated with settlement: the higher the

level of civic and social engagement, the more litigation. Finally, illegitimacy rates are

strongly positively associated with settlement: the more intact the families (the lower the

illegitimacy rate), the more litigation.

Divorces and Traffic Accidents in Japan 67

I report the analogous results from factor analysis in Regression (3) of Table 5. As

before, the coefficient on the first factor is negative and significant: the greater the level of

social capital, the lower the settlement rate. All other coefficients—including the coefficient

on attorneys per capita and on the economic factor—are insignificant.

IV. Traffic Accidents

A. The Question

Traffic accidents present an alternative database on which to test the effect of social capital

on out-of-court settlement. Through the mid-1990s (but no longer), court statistics detailed

the number of suits arising from traffic accidents. The police detail the number of fatal

(and other) traffic accidents. In Section III, I asked whether social capital facilitated the

out-of-court resolution of divorces. In this section, I ask whether it facilitates the out-of-

court settlement of serious traffic accidents.

Much as I divided the number of settled divorces by total divorces, for Table 9 I divide

the number of 1994 traffic suits by the number of 1993 traffic deaths (Keisatsu 1994:tab.

7-3). Many traffic suits did not involve wrongful death, of course. I use deaths as a denomi-

Table 8: The Effect of Attorneys on Divorce Settlements

Dependent Variable Settlement Rate

Savings PC −1.01 −1.51* −5.54 −2.16** 0.916 −1.71* −1.64*

(1.17) (1.77) (0.43) (2.27) (1.03) (1.91) (1.84)

Adv to high sch −0.007*** −0.008*** −0.008*** −0.007** −0.006*** −0.009*** −0.009***

(3.02) (3.24) (3.24) (2.45) (2.75) (3.21) (3.41)

Attorneys PC 14.156 13.525 13.490 3.647 6.644 10.320 6.657

(0.78) (0.68) (0.58) (0.16) (0.40) (0.44) (0.27)

Voter turnout −0.001**

(2.67)

Volunteering −0.001**

(2.02)

Unemployment 0.006

(1.29)

Crime rate 0.002

(1.38)

Illegitimacy 0.022***

(5.01)

Population growth 0.049

(0.53)

Metropolitan 0.006

(0.78)

Adj R20.38 0.34 0.31 0.28 0.54 0.26 0.26

***p<0.01; **p<0.05; *p<0.10.

Notes: Thetable shows the results of a regression of divorce settlement rates on several social capital measures and

the presence of attorneys. The regressions is two-stage least squares, with instruments as explained in the text. The

table gives the coefficient, followed by the absolute value of the tstatistic. All regressions include a constant term.

Coefficients on Savings are ×1,000,000. n=47. See text for definitions and sources of variables.

68 Ramseyer

nator to proxy for the quantity of major traffic disputes. I then regress this fraction on the

independent variables used above.12 Because the number of attorneys is endogenous to the

litigation level, I borrow from Ginsburg-Hoetker and instrument it with the number of

judicial scriveners (see Section II.G).

B. Variables

•Suits per death: The number of summary court (i.e., the small claims court) and

district court (the standard forum for trial) suits filed in 1994 in traffic accidents

12Because of the earlier years used for the dependent variable, I take the independent variables from years that are

as close as possible to 1994. As with the data used for the divorce regressions, however, not all variables are available

for all years. I take incomes in 1996, volunteering in 1991, unemployment in 1990, crimes in 1995, attorneys in 1997,

and the change in population over 1995–2010. I know of no major year-to-year changes in these variables and have

no reason to think that the relative rank of the various prefectures would have changed.

Table 9: Traffic Accident Litigation

Dependent Variable Suits per Death

Attorneys PC 3926.9*** 3656.3*** 3667.6*** 3659.3*** 3635.3*** 4069.2*** 3753.2***

(7.80) (5.45) (6.11) (4.40) (5.17) (5.87) (5.41)

Savings PC 6.35 3.44 19.5 1.82 21.5 6.83 2.24

(0.51) (0.30) (1.28) (0.16) (1.44) (0.59) (0.21)

Adv to high sch −0.022 −0.024 −0.007 −0.005 −0.005 0.042 0.008

(0.65) (0.77) (0.20) (0.16) (0.16) (1.18) (0.27)

Voter turnout −0.002

(0.32)

Volunteer work −0.006

(1.01)

Unemployment 0.068

(1.47)

Crime rate 0.020

(1.33)

Illegitimacy 0.134*

(1.79)

Population growth −0.782

(0.80)

Metropolitan 0.269**

(2.38)

Adj R20.72 0.73 0.73 0.74 0.74 0.72 0.78

***p<0.01; **p<0.05; *p<0.10.

Notes: Thetable shows the result of a regression of litigation rates in serious traffic accidents on various explanatory

variables. The regression is two-stage least squares, with instruments as explained in the text. The table gives the

coefficient, followed by the absolute value of the tstatistic. All regressions include a constant term. Coefficients on

Savings are ×1,000,000. n=47. See text for definitions and sources of variables. Because the dependent variable is the

number of 1994 lawsuits over 1993 traffic deaths, I use earlier independent variables where available: volunteering

activity from 1991, unemployment rates from 1990, Criminal Code violations from 1994, the number of attorneys

from 1997, the number of judicial scriveners from 1997, and population change over 1995–2010.

Divorces and Traffic Accidents in Japan 69

(current numbers are not available), divided by the number of fatal traffic acci-

dents in 1993. I take the data from Saiko (1994) and Keisatsu (1994).

C. Regression Results

1. Individual Proxy Variables

The results differ fundamentally from those involving divorces. In the divorce data, the

availability of an attorney made no difference. Instead, the settlement rate turned on social

capital, but in a way opposite to that of most discussions.

In the traffic accident data, litigation rates turn almost exclusively on the availability

of an attorney. In each specification in Table 9, the coefficient on the per-capita number of

attorneys is positive and significant: the greater the concentration of attorneys, the more

likely victims of traffic accidents will file suit rather than negotiate their settlements out of

court. Indeed, with a tstatistic ranging from 4.40 to 7.80, the coefficient is massively

significant.

The correlation coefficient between suits and lawyers is 0.86, significant at more than

the 0.1 percent level—as Figure 4 shows (Ramseyer 2012:tab. 10). Tokyo and Osaka have

the highest concentration of attorneys, of course. Drop those two and the correlation

coefficient falls, but only to 0.64 and remains significant at more than the 0.1 percent level.

Litigation rates also correlate with community income. Litigation rates are signifi-

cantly correlated with incomes and unemployment rates, and somewhat correlated to

savings (Ramseyer 2012:tab. 10). That significance disappears in the Table 9 regressions

only because attorneys work in wealthier communities.

Other than the coefficient on attorneys PC, most of the coefficients on the inde-

pendent variables in Table 9 are insignificant. The coefficient on illegitimacy is signifi-

cant at the 10 percent level: accident victims in prefectures with high illegitimacy rates are

Figure 4: Traffic accident litigation and attorneys.

0

0.5

1

1.5

2

2.5

0 0.0002 0.0004 0.0006

Suits per Death

Attorneys PC

Suits per Death

Suits per Death

Notes: Thefigure shows the positive association between a community’s litigation rate in traffic accidents, and the presence

of an attorney. The two distant prefectures are Osaka and Tokyo.See text for sources.

70 Ramseyer

more likely to sue (exactly as Lieberman and Putnam predict). The coefficient on

metropolitan is also significant and positive: accident victims in urban prefectures are

more likely to sue (again, as Putnam predicts). All other proxies for social capital are

insignificant.

2. Factor Analysis

In Regression (4) of Table 5, I report the results of a regression on the synthetic factor

variables. The results largely confirm the results from direct regressions on the proxies

themselves. People are more likely to sue in the cities. People are more likely to sue where

attorneys are common. And levels of social capital have no effect.

D. Discussion

Divorce and traffic disputes differ in the extent to which the availability of an attorney

matters. In divorce disputes, whether a couple litigates or settles has no relation to whether

an attorney is available; with traffic disputes, it does. Probably, the difference stems from the

extent to which people handle their quarrels pro se. Probably, most people handle their

divorces themselves regardless of whether attorneys have offices in their city. By contrast,

many traffic victims probably do retain a lawyer after a major accident.

This contrast between the divorce and traffic results suggests a simple moral: people

handle different kinds of disputes differently. Whether they live in a community with many

lawyers has no bearing on whether they sue or settle their divorce; it does affect whether

they sue or settle their claim over a traffic accident.

On social capital, however, the contrast largely disappears. People in communities

with high levels of social capital are not more likely to settle their divorces. Bizarrely, they

actually seem less likely to settle them. People in communities with high levels of social

capital are also unlikely to settle their traffic disputes (though illegitimacy rates and the

metropolitan dummy do generate significant results). If social capital promotes out-of-court

settlement, the Japanese data do not show it.

V. Conclusions

Scholars commonly define a “litigation rate” as the number of suits per capita. As others

(e.g., Felstiner et al. 1981) have shown, they implicitly conflate two questions: (1) How

common are the disputes? and (2) How often do people litigate those disputes? In this

article, I use data on Japanese divorces and traffic accidents to disentangle these questions.

The frequency with which couples divorce provides evidence on Question (1); whether the

parties involved litigate the divorce or accident provides evidence on Question (2).

Given the absence of micro-level time-series data, my results are only suggestive. I

find, however, that couples in poorer communities are more likely to divorce. Couples in

communities with low levels of social capital are more likely to divorce. They are not more

likely to litigate those disputes, and the accessibility of an attorney does not correlate with

that decision to litigate.

Divorces and Traffic Accidents in Japan 71

The data on traffic accidents suggest that these conclusions do not necessarily gen-

eralize. In disputes over serious accidents, the accessibility of an attorney matters crucially:

the more attorneys per capita, the higher the litigation rate. Social capital, however, is

largely uncorrelated with litigation rates.

References

Akerlof, George (1976) “The Economics of Caste and of the Rat Race and Other Woeful Tales,” 90 Q.

J. of Economics 599.

—— (1980) “A Theory of Social Custom, of Which Unemployment May Be One Consequence,” 94 Q.

J. of Economics 749.

Becker, Gary S. (1991) A Treatise on the Family: Enlarged Edition. Cambridge, MA: Harvard Univ. Press.

Bryant, Taimie L. (1984) “Marital Dissolution in Japan: Legal Obstacles and Their Impact,” 17 Law in

Japan 73.

Bunka cho (2008) Shukyo nenkan, Heisei 20 nenban [Religion Annual, 2008]. Tokyo: K.K. Gyosei (pub.

2009).

Centers for Disease Control and Prevention (2012) Data available at: http://www.cdc.gov/nchs/

nvss.htm.

Eisenberg, Theodore, Sital Kalantry, & Nick Robinson (2012) “Litigation as a Measure of Well-Being,”

SSRN: 2036194.

Ellickson, Robert C. (1991) Order Without Law: How Neighbors Settle Disputes. Cambridge, MA: Harvard

Univ. Press.

—— (1998) “Law and Economics Discovers Social Norms,” 27 J. of Legal Studies 537.

Felstiner, William, Richard Abel, & Austin Sarat (1981) “The Emergence and Transformation of

Disputes: Naming, Blaming, Claiming,” 15 Law & Society Rev. 631.

Gekkyu saachi (2009) Zenkoku kinzoku nensu rankingu [Job Tenure Rankings for the Country]. Available at

http://www.ge9.biz/local/kinzoku.html.

Ginsburg, Tom, & Glenn Hoetker (2006) “The Unreluctant Litigant? An Empirical Analysis of Japan’s

Turn to Litigation,” 35 J. of Legal Studies 31.

Hata, Ikuhito (2009) Okinawasen “shudan jiketsu” no nazo to shinjitsu [The Mystery and Truth of “Mass

Suicide” During the Battle of Okinawa]. Tokyo: PHP kenkyu jo.

Homu sho (2011) Hanzai hakusho [Crimes White Paper]. Tokyo: Homu sho.

Hook, Glenn D. (2003) “Responding to Globalization: Okinawa’s Free Trade Zone in Microregional

Context,” in G. D. Hook & R. Siddle, eds., Japan and Okinawa: Structure and Subjectivity,p.39.

London: RoutledgeCurzon.

Jacob, Herbert (1992) “The Elusive Shadow of the Law,” 26 Law & Society Rev. 565.

Katz, Avery (1996) “Taking Private Ordering Seriously,” 144 Univ. of Pennsylvania Law Rev. 1745.

Keisatsu cho, ed. (1994) Keisatsu hakusho. Tokyo: Keisatsu cho.

Kosei rodo sho (2008) Jinko ninshin chusetsu kensu [Number of Abortions]. Available at: http://

www.mhlw.go.jp/toukei/saikin/hw/eisei/07/toukei10.html.

—— (2010a) Jinko dotai tokei [Vital Statistics of Japan]. Available at: http://www.e-stat.go.jp/SG1/

estat/GL08020103.do?_toGL08020103_&listID=000001071104.

—— (2010b) “Shussei ni kansuru tokei” no gaikyo [Outline of “Statistics Relating to Birth”], press

release, December 9.

Landes, William M. (1971) “An Economic Analysis of the Courts,” 14 J. of Law & Economics 61.

Lieberman, Jethro K. (1983) The Litigious Society. Basic Books.

Malinowski, Bronislav (1962) Sex, Culture, and Myth. New York: Harcourt Trade Pub.

Mnookin, Robert H., & Lewis Kornhauser (1979) “Bargaining in the Shadow of the Law: The Case of

Divorce,” 88 Yale Law J. 950.

Monbu kagaku sho (2005) Shakai kyoiku chosa [Survey of Social Education]. Tokyo: Monbu kagaku sho.

72 Ramseyer

—— (2008) Chiho kyokuikuhi chosa hokokusho [Report of Survey on Local Educational Expenses]. Tokyo:

Monbu kagaku sho.

Murray, Charles (2012) Coming Apart: The State of White America, 1960–2010. New York: Crown Forum.

Nakazato, Minoru, J. Mark Ramseyer, & Eric B. Rasmusen (2009) “Public and Private Firm Compen-

sation Compared: Evidence from Japanese Tax Returns,” 25 Korean Economic Rev. 5.

—— (2010) “The Industrial Organization of the Japanese Bar,” 7 J. of Empirical Legal Studies 460.

—— (2011) “Executive Compensation in Japan: Estimating Levels and Determinants from Tax

Records,” 20 J. of Economics & Management Strategy 843.

Nenshu rabo (2012) Todofukenbetsu nenshu rankingu [Ranking of Annual Income, by Prefecture].

Available at http://nensyu-labo.com/2nd_ken_ranking.htm (for 2008 figures).

Nihon ABC Kyokai (2011) Shinbun hakkosha repoto fukyu ritsu [Diffusion Rate, Newspaper Publisher Report],

July–Dec. Available at: adv.yomiuri.co.jp/yomiuri/busu/busu01b.html.

Nihon bengoshi rengo kai (2011) Bengoshi hakusho [Attorney White Paper]. Tokyo: Nihon bengoshi rengo

kai.

Nihon shiho shoshi rengo kai (2012) Nihon shiho shoshi rengo kai [Japan Federation of Judicial Scriveners]

(home page). Available at: http://www.shiho-shoshi.or.jp/association/shiho_shoshi_list.php.

Ostrom, Elinor (1998) “A Behavioral Approach to the Rational Choice Theory of Collective Action,” 91

American Political Science Rev. 1.

Posner, Richard A. (1973) “An Economic Approach to Legal Procedure and Judicial Administration,”

2J. of Legal Studies 399.

—— (1997) “Social Norms and the Law: An Economic Approach,” 87 American Economic Rev. 365.

—— (1998) “Social Norms, Social Meaning, and Economic Analysis of Law: A Comment,” 27 J. of Legal

Studies 553.

Posner, Richard A., & Eric B. Rasmusen (1999) “Creating and Enforcing Norms, with Special Refer-

ence to Sanctions,” 19 International Rev. of Law & Economics 369.

Putnam, Robert D. (2000) Bowling Alone: The Collapse and Revival of American Community. New York:

Simon & Schuster.

Ramseyer, J. Mark (2009) “The Effect of Cost Suppression Under Universal Health Insurance on the

Allocation of Talent and the Development of Expertise: Cosmetic Surgery in Japan,” 52 J. of Law

& Economics 497.

—— (2012) Litigation and Social Capital: Divorces and Traffic Accidents in Japan. Available at: http://

papers.ssrn.com/sol3/papers.cfm?abstract_id=2138229.

Saiko saibansho, ed. (1994) Shiho tokei nempo: 1 Minji, gyosei hen [Annual Report of Judicial Statistics: 1 Civil

& Administrative Cases]. Tokyo: Hoso kai.

Somu sho (2003) Senkyo kanren shiryo [Materials Related to Elections]. Available at: http://

www.soumu.go.jp/senkyo/senkyo_s/data/shugiin44/index.html.

—— (2004) Zenkoku shohi jittai chosa [Survey of National Consumption]. Tokyo: Somusho.

—— (2006) Shakai seikatsu kihon chosa [Basic Survey of Social Life]. Available at: http://www.stat.go.jp/

data/shakai/2011/trivia/index.htm#table3.

—— (2012) Rodoryoku chosa [Survey of Labor Force]. Available at: http://www.stat.go.jp/data/roudou/

pref/index.htm (for unemp).

Stata (2005) Stata Multivariate Statistics Reference Manual: Release 9. College Station, TX: Stata Press.

Toba, Ken (2005) Nihonjin no heikinchi [Average Values for Japanese]. Tokyo: Seikatsu joho sentaa.

Tokyo shoko risaachi (2004) Zenkoku kogaku nozeisha meibo [National Registry of High-Income Taxpayers],

CD-ROM version. Tokyo: Tokyo shoko risaachi.

Divorces and Traffic Accidents in Japan 73