Notre Dame Law School Notre Dame Law School
NDLScholarship NDLScholarship
Journal Articles Publications
1998
No-Fault Laws and At-Fault People No-Fault Laws and At-Fault People
Margaret F. Brinig
Notre Dame Law School
F. H. Buckley
Follow this and additional works at: https://scholarship.law.nd.edu/law_faculty_scholarship
Part of the Family Law Commons
Recommended Citation Recommended Citation
Margaret F. Brinig & F. H. Buckley,
No-Fault Laws and At-Fault People
, 18 Int'l Rev. of L. & Econ. 325
(1998).
Available at: https://scholarship.law.nd.edu/law_faculty_scholarship/680
This Article is brought to you for free and open access by the Publications at NDLScholarship. It has been accepted
for inclusion in Journal Articles by an authorized administrator of NDLScholarship. For more information, please
contact [email protected].
No-Fault Laws and At-Fault People
MARGARET F. BRINIG
and
F.H. B
UCKLEY
George Mason University School of Law, Fairfax, Virginia, USA
Absent transaction costs, the Coase Theorem suggests that divorce reform would
work no change in the frequency of divorce but perhaps would alter the distribution of
marital wealth. However, divorce does involve substantial process costs, which no-fault
lowered. This paper explores the question of what happened to state divorce rates
because of the legal changes wrought by the family law revolution that began in the
1970s, isolating the effect of the legal variable from other demographic and social
factors that might also explain the variation in divorce rates across states and across
time. © 1998 by Elsevier Science Inc.
I. Introduction
Divorce law reform is in the air. For the first time in memory, moreover, reformers are
seeking to strengthen barriers to divorce. In the past, divorce reform always meant
easier divorces. But today, reformers seek the repeal of no-fault divorce laws, which they
say have weakened the family.
1
Between 1969 and 1985, every state liberalized its divorce laws. The change was largely
complete by 1979, when only two states required proof of fault before granting the
divorce. Divorce rates almost doubled during this period, as may be seen in Figure 1.
This was unsurprising, because the move to no-fault laws reduced the cost of divorce
and sapped spousal incentives to invest in their marriage. However, the claim that
no-fault divorce laws result in a long-term increase in divorce rates has been challenged.
Elizabeth Peters has argued that, because of Coasian bargaining, divorce levels should
We acknowledge the support of the George Mason University Law and Economics Center, and the helpful
comments of Doug Allen, Allen Parkman, Eric Posner, Eric Rasmusen, anonymous referees of this Journal, and the
participants at annual meetings of the American Law and Economics Association and the American Association of Law
Schools (AALS).
Article accepted March 1997.
1
Wark, John T. “Finding Fault.” Detroit News, February 13, 1996.
International Review of Law and Economics 18:325–340, 1998
© 1998 by Elsevier Science Inc. 0144-8188/98/$19.00
655 Avenue of the Americas, New York, NY 10010 PII S0144-8188(98)00008-8
not be affected by the legal regime.
2
In support of this position, she conducted a panel
study and reported that no-fault laws were not correlated with higher divorce rates.
This article takes issue with these claims. We suggest that divorce levels are dependent
on the legal regime and that they are lower in states that penalize marital fault. Our
results differ from those of Peters primarily because of differences in our definitions of
“no fault.” No fault for Peters meant that fault was irrelevant in the dissolution pro-
ceedings, whether or not it affected the division of assets or alimony award. But this is
short sighted, because one would expect less fault and fewer divorces when fault bears
a financial penalty. Accordingly, we define “no fault” to mean that fault is irrelevant at
both dissolution and at financial settlement.
Our results suggest a policy response to increased divorce levels. Social conservatives
argue persuasively that increased divorce levels have harmed women and children and
coarsened civil society. Their efforts to condition the grant of the divorce on a finding
2
Peters, H. Elizabeth. (1986). “Marriage and Divorce: Informational Constraints and Private Contracting.” American
Economic Review 76:437.
FIG. 1. U.S. divorce rates.
326 No fault laws and at-fault people
of fault might nonetheless fail, as fresh start norms are deeply ingrained in America.
3
A
more promising strategy, with broader appeal, might seek to reintroduce fault into
property awards, while retaining no-fault standards in the dissolution proceedings. Our
findings suggest that such a change would effectively reduce divorce rates.
II. Fault and No Fault
No state today requires fault for the dissolution of a marriage. Many states have adopted
the no-fault standards of the Uniform Marriage and Divorce Act,
4
in which the only
ground for granting a dissolution is that the relationship is “irretrievably broken.” For
example, the Colorado statute provides that the “petition...shall allege that the
marriage is irretrievably broken,” and that “[d]efenses to divorce and legal separa-
tion...,including but not limited to condonation, connivance, collusion, recrimina-
tion, insanity, and lapse of time, are hereby abolished.”
5
Several empirical studies report a short-term increase in divorce rates after such
statutes were enacted.
6
This result is what one would have expected. Because no-fault
laws reduced the cost of divorce,
7
some couples whose marriages were failing postponed
divorce until the new law took effect. The result was a blockage in the divorce pipeline
just before enactment, and a surge immediately afterwards.
There is less agreement about the long-term effects of no-fault divorce laws. The most
well-known paper, by Elizabeth Peters,
8
reported that no-fault laws had virtually no
influence on the probability of a couple’s divorcing between 1975 and 1977. More
recently, other economists have criticized the Peters study. Douglas Allen reported that
Peters’ no-fault predictor was significantly and positively correlated with divorce when
he omitted data from three states that moved to no-fault during the period of the Peters
study, and which Peters had labeled as fault states.
9
In response, Peters faulted Allen for
omitting regional variables and a predictor that Allen had employed as a proxy for
fixed-state effects.
10
A study by Martin Zelder reported that a no-fault variable was not
correlated with increased divorce levels.
11
However, when the variable was multiplied by
a measure of expenditures on children, the new interactor was associated significantly
with increased divorce levels. From this, Zelder concluded that divorce law irrelevance
does not hold when the parties have children.
12
3
The popular objections to fault grounds, including the distasteful and collusive process of “staged” grounds, are
discussed in Herbert Jacobs, The Silent Revolution, University of Chicago Press, 1988.
4
9A U.L.A. § 305.
5
Colo. Rev. Stat. § 14-10-107 (1995).
6
See Marvell, Thomas B., (1989). “Divorce Rates and the Fault Requirement.” Law & Society Review 23:543, 544; Paul
A. Nakonezy, Robert D. Shull, and Joseph Lee Rodgers. (1995). “The Effect of No-Fault Divorce Law on the Divorce
Rates Across the 50 States and Its Relation to Income, Education and Religiosity.” Journal of Marriage and the Family
57:477.
7
Mnookin, Robert H., and Lewis Kornhauser. (1979). “Bargaining in the Shadow of the Law: The Case of Divorce.”
Yale Law Journal 88:950.
8
Peters, supra note 2.
9
Allen, Douglas W., (1992). “Marriage and Divorce: A Comment.” American Economic Review 82:297. See also, Allen
Parkman, (1992). “Unilateral Divorce and the Labor Force Participation Rate of Married Women, Revisited.” American
Economic Review 82:671.
10
Peters, H. Elizabeth. “Marriage and Divorce: Reply.” American Economic Review 82:686.
11
“Inefficient Dissolutions as a Consequence of Public Goods: The Case of No-Fault Divorce.” Journal of Legal Studies
22:503.
12
About 60% of married couples have children. Elizabeth Scott, (1990). “Rational Decisionmaking about Marriage
and Divorce.” Virginia Law Review 76:9, 94, at note 8.
327BRINIG AND BUCKLEY
All of these studies employed a no-fault variable that ignored whether or not fault
affected the financial award. States were labeled as no-fault even though fault was
penalized, which narrowed the difference between the divorce rates in fault and
no-fault states. This plausibly explains the failure to detect significantly lower divorce
levels in fault states.
13
Accordingly, we define “no fault” to mean that fault is irrelevant
at both the dissolution and financial settlement states. This permits a sharper picture of
fault requirements than the above studies. Table 1 gives our list of fault and no-fault
states, together with the year that no-fault laws were adopted.
As we define fault, a law is no-fault if matrimonial fault is irrelevant for dissolution, for
the division of assets, and for alimony. For example, we view the Montana statute as a
true no-fault law. Because the state has adopted the Uniform Marriage and Divorce Act
state, fault is irrelevant at dissolution. In addition, the Montana statute provides that the
“maintenance order shall be in such amounts...as the court deems just, without
regard to marital misconduct, and after considering all relevant facts.”
14
By contrast, we
regard the Virginia law as a fault statute. Although a Virginia divorce may be granted on
either fault or no-fault grounds, in awarding alimony the judge is enjoined to “consider
the circumstances and factors which contributed to the dissolution of the marriage,
specifically including adultery and any other ground for divorce.”
15
Western states have
generally adopted no-fault standards, whereas eastern and southern states continue to
penalize fault (see Figure 2).
The claim that divorce rates are independent of the fault regimes in divorce asks one
to look at the divorce decision from an ex post perspective, at the time of divorce, when
sentiments have hardened and the parties have determined their course of action. At
that point, it is said, the parties can always bargain around legal rules, as the Coase
theorem suggests. In a fault regime, when a husband wants a divorce more than his wife
wants to preserve the marriage, he can bribe her to consent to the divorce; in a no-fault
regime, when a wife wants to preserve a marriage more than her husband wants a
divorce, she can bribe him to stay married. Assume, for example, that a husband would
pay up to $300 for a divorce, whereas his wife would pay up to $100 to preserve the
marriage. They will then divorce, whatever the legal regime. In a fault regime, the wife
will agree to sue for (or consent to) divorce on fault grounds for a payoff of between
$100 and $300. In a no-fault state, the husband will unilaterally initiate divorce pro-
ceedings, rejecting the wife’s offer of $100 to remain married.
There are, however, two reasons why Coasean irrelevance might not hold. First, the
wife might be unwilling to bribe the husband to stay married because a promise not to
13
We preferred our definition of fault to that of the 1996 American Law Institute (ALI) Principles of Family
Dissolution—Analysis and Recommendations 60 (Tentative Draft No. 2). Our list of fault states differs from that of the
ALI Report in three respects. First, unlike the ALI, we label a state as a fault state if divorces may be obtained on fault
grounds, because this permits the innocent party to introduce marital misbehavior into settlement negotiations. By
contrast, the ALI looks only to fault as it involves alimony or property distribution. Id. at 17. This difference affects three
states. See Alaska Rev. Stat. § 25.24.050 (1995); Maine Rev. Stat. Ann. tit 19, § 691 (1995); and N.M. Stat. Ann. § 40-4-1
(1996). Second, our list is based on the 1979 to 1990 period, whereas the ALI list is from 1996. Fault is penalized in
fewer states today than during the period of our study. Delaware, Illinois, and Kentucky all moved to no-fault between
1979 and 1990. Third, we excluded Nevada from our study as an outlier because it is a national divorce haven.
14
Mont. Code Anno. § 40-4-203 (1995). The “relevant facts” might refer to the parties’ age or earning capacity but
not to matrimonial fault.
15
Va. Code Ann. § 20-107.1 (1995). Labeling the law as fault or no-fault sometimes required an examination of case
law to see whether fault was penalized in states where the parties were given the option of seeking a divorce on either
fault or no-fault grounds.
328 No fault laws and at-fault people
seek a divorce is unenforceable at law.
16
The husband might pocket the bribe and bring
a petition the next day in a no-fault regime. Second, the preferences of the parties
might be endogenous and affected by the divorce law regime. The amount that the wife
might be willing to pay in a no-fault regime to maintain the marriage (in dollars or
complaint behavior) might be less than the minimum amount that she would accept to
surrender her veto rights in a fault regime. The economist refers to this difference as an
endowment or income effect, inasmuch as it reflects the shift in her budget line when
legal rules change. Preferences also may shift because of the ex ante incentive effects of
legal rules on behavior and preferences. When matrimonial fault is not penalized, there
16
Towles v. Towles, 256 S.C. 307, 182 S.E.2d 53 (1971).
TABLE 1. Year of adoption of no-fault regimes
Alabama 0 Montana 1975
Alaska 0 Nebraska 1972
Arizona 1974 Nevada 0
Arkansas 0 New Hampshire 0
California 1969 New Jersey 0
Colorado 1971 New Mexico 0
Connecticut 0 New York 0
Delaware 1979 North Carolina 0
Florida 1978 North Dakota 0
Georgia 0 Ohio 0
Hawaii 1972 Oklahoma§ 1975
Idaho 0 Oregon 1971
Illinois* 1984 Pennsylvania 0
Indiana 1973 Rhode Island 0
Iowa 1971 South Carolina 0
Kansas† 0 South Dakota 0
Kentucky‡ 1987 Tennessee 0
Louisiana 0 Texas 0
Maine 0 Utah 0
Maryland 0 Vermont 0
Massachusetts 0 Virginia 0
Michigan 0 Washington 1973
Minnesota 1974 West Virginia 0
Mississippi 0 Wisconsini 1977
Missouri 0 Wyoming 0
Notes: States that have not enacted law making fault irrelevant for all
purposes are denoted by a 0.
*Harambasic v. Harambasic, 370 NE2d 1251, 1255 (Ill App 1977) (eliminating
fault from grant of alimony); Ill Comp Stat Ann § 5/501 (adding irreconcilable
differences as ground for divorce).
†In re Marriage of Sommers, 792 P2d 1005, 1008 (Kan 1990) (defining divorce
statue as including fault in the “breach of marital duty” section).
‡Moss v. Moss. 639 S.W.2d 370, 371 (Ky Ct App 1987) (stating that alimony
was to be granted without regard to fault).
§Smith v. Smith, 847 P2d 827, 829 (Ok 1993) (stating that fault has not been
used in the granting of alimony since passage of the 1975 law).
iDixon v. Dixon, 319 NW2d 846, 849-50 (Wis 1982) (finding fault irrelevant
in the granting of alimony since passage of no-fault divorce law).
329BRINIG AND BUCKLEY
will be more of it, and more straying or outraged spouses who seek a divorce.
17
There
will be more grievous faults, and trivial breaches will be forgiven less readily. The slight
offense that is quickly forgotten in a fault regime, where divorce is not seen as an option,
might be felt to oppress in a no-fault one.
18
With a greater probability of divorce, the
parties also will invest less in marriage-specific assets such as children, and this will
further increase divorce levels.
19
As well, the social stigma of divorce might weaken
when no-fault laws are passed and divorce levels increase. When relaxed social norms
are internalized, the transgressor will no longer be policed by a sense of guilt.
Ex ante incentive effects might in theory be addressed through private fault barriers
in prenuptial agreements in no-fault regimes. For example, a wife might bargain for
one-quarter of her husband’s future earnings on divorce if neither are at fault, but for
three-quarters if he is guilty of a matrimonial offense and she is not. However, we are
unaware of any case in which the spouses have actually bargained for a private fault
regime in a no-fault state. We can think of at least two reasons for this. First, such a
bargain might be held illegal, insofar as it conflicts with the no-fault statute. Would it be
upheld in Montana, for example, where the statute explicitly provides that the main-
tenance order shall be made “without regard to marital misconduct”? Even where the
17
Richard A. Posner, (1992). Sex and Reason, Cambridge, MA: Harvard University Press, p. 249; see also Margaret F.
Brinig, and Steven M. Crafton, (1994). “Marriage and Opportunism.” Journal Legal Studies 23:869, 873.
18
Scott J. South, (1985). “Economic Conditions and the Divorce Rate: A Time-Series Analysis of the Postwar United
States.” Journal of Marriage and Family 47:31. Writing much earlier, this is the sentiment of David Hume, (1987). Of
polygamy and divorces. In Essays: Moral, Political and Literary, Eugene Miller, ed. 181–190. Indianopolis, IN: Liberty
Classics. The fact that law does have an impact on behavior, particularly moral behavior, is the focus of Carl E.
Schneider, (1992). “The Channelling Function in Family Law.” Hofstra Law Review 20:495.
19
Brinig and Crafton, supra note 17.
FIG. 2. Fault and no-fault states.
330 No fault laws and at-fault people
conflict with the statute is not so clear, the ability of the parties to stipulate for a
financial penalty on divorce is clouded with uncertainty.
20
Second, a party might not
seek to bargain for stiff fault penalties because this might send a signal that she is
mercenary and might cost her the marriage.
21
III. Our Model
This section discusses the results of a regression analysis of the determinants of state
divorce levels. Our principal finding is that divorce levels are positively and significantly
correlated with state laws that do not penalize marital misbehavior at the time of
divorce.
Our dependent variable DIVORCE is the per capita state divorce rate. To examine the
effect of state divorce laws on DIVORCE levels in State
i
, we estimated the following
equation:
DIVORCE
it
5 a
0
1 b
1
LAW
it21
1 b
2
UNEMPLOYMENT
it21
1 b
3
EMPLOYMENT GROWTH
it21
1 b
4
YEAR 1 b
5
ENTRY
i
1 b
6
METRO
it-1
1 b
7
INSURE
it21
1 b
8
CATHOLIC
it21
1 e
i
where e
i
is an error term, and where the variables are as defined in Table 2. Summary
statistics are given in Table 3. Divorce levels from 1980 to 1991 were regressed on 1979
to 1990 independent variables. Unless otherwise mentioned, the source for our data was
the Statistical Abstract of the United States.
All non-dummy variables were transformed into their natural logarithmic form, after
20
Sanders v. Sanders, 288 S.W.2d 473 (Tenn. 1955) (suggesting that a financial penalty for divorce in antenuptial
agreement would be contrary to public policy if the divorce suit was prosecuted in good faith and on reasonable
grounds); Norris v. Norris, 174 N.W.2d 368 (Iowa 1970) (impeaching an antenuptial fault clause). See further Theodore
F. Haas, (1988). “Rationality and Enforceability of Contractual Restrictions on Divorce.” North Carolina Law Review
66:879 (arguing for the enforceability of financial fetters on divorce in antenuptial agreements); Elizabeth Scott,
(1994). “Rehabilitating Liberalism in Modern Divorce.” Utah Law Review 687:722.
21
The signaling problem would persist if the parties were asked to elect between a fault and a no-fault regime, as
Jeffrey Stake has suggested. Jeffrey E. Stake, (1992). “Mandatory Planning for Divorce.” Vanderbilt Law Review 45:397.
In the past, the signaling problem was addressed through the appointment of parents to bargain on behalf of their
engaged children.
Apart from illegality and signaling problems, the myopic might fail to provide for the consequences of divorce in
an antenuptial agreement because they ascribe too low a probability to divorce. See Scott, supra note 20, at 722.
TABLE 2. Definition of variables
Divorce rate Divorces divided by 1000 population
No-fault divorce 1 5 Fault irrelevant in divorce and property settlements; 0 otherwise
Peters unilateral
divorce
1 5 Statute allows unilateral divorce; 0 5 statute allows mutual consent
divorce only
Unemployment Average of monthly unemployment figures
Employment growth Percentage of yearly change in non-farm employment
Metropolitan Percent of population living in metropolitan areas
Insurance Total dollar amount of life insurance in the state, divided by state income
Entry Year of admittance of state into the Union less 1788 (original states as 1)
Marriage rate Marriages divided by 1000 population
331BRINIG AND BUCKLEY
we determined that this was appropriate through a Box-Cox test on untransformed
variables.
22
The Dependent Variable
Our dependent variable, DIVORCE, represents total state divorces per year divided by
state population. This is not the only way to measure divorce levels. Divorces also might
be computed per married couple. Per capita divorce rates may be low because marriage
rates are low, and not because married people are more faithful or happier in that
state.
23
The divorce rate will be zero in a state where no one marries. Because of this,
we re-estimated divorce levels on a per couple basis. Because the results were so similar
to the per capita figures,
24
however, we omit them. We also employed marriage levels as
an endogenous variable in the two-stage least square estimations that we describe below.
Divorce rates are far higher in Nevada than in other states.
25
Unlike Peters and Allen,
we therefore excluded Nevada observations from our study. We were left with 49 states,
over a 12-year period, for a total of 588 observations.
Independent Variables
We employed two different legal predictors of divorce. In the first two columns of Table
4, the NO-FAULT estimator is a dummy variable that takes the value of 1 if the state has
a no-fault divorce law, as we have defined it, and takes 0 otherwise. In the third and
fourth columns, the PETERS estimator is a dummy variable that takes the value of 1 if
the state has a no-fault divorce law, as Peters has defined it, and takes 0 otherwise.
26
22
Our Box-Cox ls on our fixed-state effects models ranged from 0.38 to 0.43, and on our non-fixed-state effects
models ranged from 0.24 to 0.36. Box-Cox transformations are discussed in George C. Judge, R. Carter Hill, William
E. Griffiths, Helmut Lu¨tkepohl, and Tsoung-Chao Lee, (1987). Introduction to the Theory and Practice of Econometrics. 2nd
ed. New York: Wiley, pp. 555–556.
23
We obtained the number of married couples in each state for 1980 and 1990 from the decennial censuses. For
intervening years, we multiplied the state population by the married couple ratio, which we constructed by assuming
an equal change each year.
24
Table available from authors.
25
From 1973 to 1991, Nevada’s average annual divorce rate was 15.47 per 1000, about three times higher than the
mean of 5.08 when Nevada was excluded. The next highest rate was approximately 8.
26
See supra text in notes 14 and 15.
TABLE 3. Summary statistics
Variable Mean Standard Deviation Minimum Maximum
DIVORCE 5.0259 1.7156 2.2000 9.1000
UNEMPLOYMENT
21
6.7985 5.4724 1.7000 18.000
EMPLOYMENT GROWTH
21
101.71 3.0070 92.946 113.68
METRO
21
62.451 22.276 15.000 100.00
CATHOLIC
21
19.370 13.494 1.6000 65.800
INSURE
21
2.6185 0.20269 1.3966 4.7620
MARRIED COUPLES
21
10.213 2.0790 6.8113 17.353
WORKING WIVES
21
53.910 38.901 36.600 64.900
INCOME
21
1237.2 238.65 823.97 2252.3
332 No fault laws and at-fault people
Because we expected higher divorce levels in no-fault states, our model predicted that
both coefficients would be positive, particularly NO-FAULT.
The two legal variables should be seen as proxies for the financial sanctions that
courts actually impose. Ultimately, what counts is the value of the expected sanction for
marital fault, and not the statutory regime. Moreover, the statutory regime is not always
a perfect proxy for the financial sanctions. In a no-fault regime, a conservative judge
might impose a financial penalty on a misbehaving spouse,
27
whereas a liberal judge in
a fault state might offer an easy absolution. The statutory variable might thus be prone
to type I and type II errors. Local legal cultures might affect financial awards in other
ways. For example, awards generally might be higher in one state for in all civil and
marital actions. Because of this, we would have relied on the actual financial payouts
had they been available.
28
Nevertheless, our assumption that the financial penalty will
27
In egregious cases, issues of fault may surface in tort actions See Brinig and Crafton, supra note 17, at 893 and note
108. Fault might also result in an annulment. See Brinig, Margaret F., and Michael V. Alexeev, (1994). “Fraud in
Courtship.” European Journal of Law and Economics 2:45.
28
The evidence from panel studies indicates that payouts are higher in fault states (as we have defined that term).
See Margaret F. Brinig, and Michael V. Alexeev, (1993). “Trading at Divorce: Preferences, Legal Rules and Transaction
Costs.” Ohio State Journal of Dispute Resolution 8:279; Marsha Garrison, (1991). “Good Intentions Gone Awry: The Impact
of New York’s Equitable Distribution Law on Divorce Outcomes.” Brooklyn Law Review 57:621; Michael Kelly and Greer
Litton Fox, (1993). “Determinants of Alimony Awards.” Syracuse Law Review 44:641; Elisabeth Landes, (1978). “The
TABLE 4. The determinants of per capita divorce rates
Variable 1234
NO-FAULT
21
0.13694
(4.313)*
0.15812
(5.209)*
PETERS
21
0.045626
(1.379)
0.066747
(2.067)*
UNEMPLOYMENT
21
20.028642
(24.182)
20.027235
(23.984)*
EMPLOYMENT GROWTH
21
0.19347
(4.681)*
0.21097
(5.120)*
YEAR 20.013892
(22.656)*
20.020128
(23.747)*
20.012452
(22.372)*
20.018924
(23.480)*
ENTRY 0.17172
(3.674)*
0.21891
(5.547)*
0.17354
(3.709)*
0.23417
(5.821)*
METRO
21
0.037456
(1.444)
0.15786
(4.397)*
0.035274
(1.372)
0.16097
(4.372)*
CATHOLIC
21
0.061140
(2.398)*
0.10640
(4.320)*
0.066332
(2.589)*
0.11348
(4.594)*
INSURE
21
20.23086
(27.952)*
20.21812
(27.207)*
20.23694
(28.099)*
20.22427
(27.292)*
Sum of squared errors 581.42 585.43 580.98 584.79
Standard error 1.0464 1.0480 1.0460 1.0475
Buse R
2
(1979) 0.4675 0.4526 .4585 0.4388
Log likelihood 940.121 932.555 936.305 928.414
Note: Estimated regression coefficients and Kmenta pooling with fixed state effects.
*significant at .05 level
333BRINIG AND BUCKLEY
be higher in fault states is not unreasonable. If conservative and liberal (and high- and
low-payout) judges are normally distributed among fault and no-fault states, we would
still expect a higher mean payout in fault states, so long as some judges pay attention to
statutory standards. And if judges are not normally distributed, we would expect to find
more conservative ones in fault states, if the political sentiment on the bench mirrors
popular political sentiment in the state.
The social capital theorist expects social variables to be highly correlated with eco-
nomic and legal variables. Social norms affect the level of material wealth, which in turn
affects social norms. Perverse laws also may subvert social norms, and perverse social
norms may subvert the enforcement and enactment of legal norms. For example, a
society with liberal social norms is more likely to enact no-fault laws, and such a society
also is more likely to have high divorce rates. Our legal variables thus might be partly
social in nature, whereas our social variables might partly be legal constructs.
29
We
address the multicollinearity problem by employing specifically social variables, in
addition to our legal variables.
As a proxy for economic growth, we employed two predictors. UNEMPLOYMENT is
the yearly average of monthly unemployment rates
30
; and EMPLOYMENT GROWTH is
the percentage of increase in nonfarm total employment from year to year. We might
expect to find higher divorce rates after economic downturns. Economic hardship
imposes strains on many marriages. A spouse might have to work harder, spending
more time away from his family, possibly even migrating to another state. In addition,
where the parties have seen their wealth disappear in a severe economic downturn, the
financial costs of a divorce might seem less troubling.
31
If one has lost nearly everything,
there comes a point when there is less need to preserve what one has by staying married.
The effect of economic predictors is ambiguous, however, because couples divorce in
both good times and bad, and social variables seem to matter more than the economic
ones. For example, divorce rates are far higher today than in the Great Depression. To
the extent that economic variables matter, volatility seems to be more important than
mean values.
32
If spouses are either much wealthier or much poorer than they expected
at marriage, divorce is more likely.
33
Divorce rates are sensitive to social norms, because social stigma may greatly increase
the cost of deviant behavior. The decline in the stigma of divorce in this century likely
explains much of the increase in divorce levels. We would also expect social sanctions
to vary from one region to another. Therefore, we employed four social predictors of
Economics of Alimony.” Journal of Legal Studies 7:35. For example, Yoram Weiss, and Robert Willis, (1993). “Transfers
Among Couples in Divorce Settlements.” Journal of Labor Economics 11:629, 656 at Table 4, show that divorced wives with
children received a mean of $9313 in no-fault states, compared to $5220 in fault states (as we define them). In most
of these studies, however, the difference in payouts is not significant.
29
The reductionist and unverifiable claim that all variables are social in nature is as unreasonable as the claim that
at the bottom all variables are proxies for more fundamental economic conditions or legal variables. For a political
explanation of when no-fault laws were introduced, employing a logit estimation technique, see Brinig and Crafton,
supra note 17.
30
This is the percentage of members of the labor force who are actively looking for employment.
31
Bumpass, Larry, Teresa Castro Martin and James Sweet, (1990). “Background and Early Marital Factors in Marital
Disruption.” Madison, WI: Center for Demography and Ecology, unpublished manuscript, pp. 10 and 16 (reporting a
two-thirds greater probability of divorce if the husband was unemployed at any time during the first year of marriage).
32
Becker, Gary S., (1991). A Treatise on the Family. Cambridge, MA: Harvard University Press, p. 339.
33
Becker, Gary, Elisabeth M. Landes, and Robert Michael, (1977). “An Economic Analysis of Marital Instability.”
Journal of Political Economy 85:1141; South, supra note 18, at 37.
334 No fault laws and at-fault people
divorce rates: ENTRY, METRO, INSURE, and CATHOLIC. We omitted racial variables,
which were insignificant in both the Peters and Allen studies.
As may be seen in Figure 3, divorce rates are higher in western states.
34
Regional
differences are long-standing,
35
because the frontier offered spouses a relatively easy
exit option from marriage.
36
In 1838, Alexis de Toqueville noted the “restless disposi-
tion and an excessive love of independence” of Americans, whose westward migration
kept “ties broken or unformed.”
37
Even today, higher divorce rates in western states
plausibly reflect a more mobile society, with a greater proportion of migrants and
weaker ties to the social and family institutions that prop up ailing marriages. Regional
differences in divorce rates also might reflect different social sanctions for divorce.
One might seek to capture the frontier effect through regional dummy variables.
However, dummy variables do not weigh the frontier effect, as the ENTRY variable does.
ENTRY represents the number of years between 1788 and the year that the state was
admitted to the Union,
38
with increasing values as one moves westward.
34
Norval D. Glenn, and Beth Ann Shelton, (1985). “Regional Differences in Divorce in the United States.” Journal
Marriage and Family 47:641; Norval D. Glenn, and Michael Supanic, (1984). “The Social and Demographic Correlates
of Divorce and Separation in the United States: An Update and Reconsideration.” Journal of Marriage and Family 47:563.
35
In 1908, the Labor Department reported that “the divorce rate increases as one goes westward.” United States
Department of Labor and Commerce, (1908). Marriage and Divorce 1867–1906, I, 14-15, (Reprinted Westport, CT,
1978).
36
Roderick Phillips, (1988). Putting Asunder. Cambridge: Cambridge University Press, p. 452.
37
De Toqueville, Alexis, (1838). Democracy in America. New York, NY: Random House, pp. 276 and 285.
38
The ENTRY variable took the value of 1 for each of the original thirteen states. An ENTRY variable was used to
predict state economic growth in Olson, Mancur. (1982). The Rise and Decline of Nations. New Haven, CT: Yale University
Press.
FIG. 3. Divorce rates from 1979 to 1991.
335BRINIG AND BUCKLEY
Urbanization always has been linked with increased divorce rates.
39
Cities offer
greater opportunities to stray and greater anonymity for transgressors. In cities, more-
over, the safety net of community and religion might be weaker than in smaller towns.
No doubt, that has always been part of the appeal of cities. Our proxy for urbanization
was the METRO variable, representing the percentage of the population living in a
metropolitan statistical area.
The link between marriage and insurance was noted by Gary Becker,
40
who argued
that marriage reduces the financial and emotional risks that singles bear.
41
In marriage,
spouses pool their fortunes by promising “to love and honor...in sickness and in
health, for better and for worse.”
42
The emotional security of marriage also will have a
higher appeal for the risk averse. No doubt, marriage “is a lottery in which there are
wondrous many blanks,” as Voltaire noted. But even in such cases, the known quantity
of one’s spouse may be preferred to the uncertainty of the next (blind) date or the
unhappiness of remaining alone.
43
Our INSURE variable, representing total life insurance in force divided by total
income for each state, serves as a proxy for differential risk aversion. We suggest that
those who are risk averse in one set of decisions will be reluctant to gamble in other sets
of choices, even where the decisions are ostensibly dissimilar. As a measure of general
risk aversion, there are few better sources of data than life insurance rates, which vary
substantially from one region to another.
44
The CATHOLIC variable represents the percentage of Catholics in the state.
45
More
than any other major religion, Catholicism refuses to recognize remarriage after di-
vorce. Divorce rates also might be lower for Catholics if they are more risk averse than
Protestants, as Max Weber famously argued.
46
However, the differences in divorce rates
are smaller than one might have thought. In America, 26% of Protestants and 23% of
Catholics have been divorced at least once.
47
39
Phillips, supra note 36.
40
Gary S. Becker, (1993). “The Economic Way of Looking at Behavior.” Journal of Political Economy 101:385. See also,
Christopher J. Bruce, (1994). “An Economic Model of Spousal Support,” Part III. Working paper, University of Calgary
Department of Economics.
41
A point noted well before Becker. See, e.g., Karl Llewellyn, (1932). “Behind the Law of Divorce,” Part 1. Columbia
Law Review 32:1281, 1290.
42
See, e.g., West Virginia Code §48-1-12b (1994), and the traditional marriage ceremony in the Book of Common
Prayer.
43
Becker, et al. supra note 33, pp. 1147–1148; Paula England and Gary Farkas, (1988). Households, Employment and
Gender. New York: Aldine Press; William Bishop, (1984). “Is He Married?: Marriage as Information.” University of Toronto
Law Journal 23:245, 249; Becker, supra note 32, at 337.
44
We re-estimated divorce levels in regressions that omitted the INSURE variable and found that little changed,
particularly with respect to our legal coefficients. Data available from authors.
45
Source: Official Catholic Directory, years 1970 to 1991, published by P.J. Kennedy & Sons in association with R.R.
Bowker, A. Reed Reference Publishing Company, New Providence, NJ.
46
Max Weber, (1930). The Protestant Ethic and the Spirit of Capitalism.
47
George Gallup Jr. and Jim Castelli, “Attitudes On Marriage Surveyed; Catholic, Protestant Divorce Rates Similar,”
Washington Post, Saturday, April 8, 1989, at C13. These figures understate the difference in divorce rates, because
Protestants are more likely to remarry and have a second divorce. Although 54% of divorced Protestants remarry, only
39% of divorced Catholics do so.
Once again, we re-estimated divorce levels in regressions that omitted the CATHOLIC variable and found that little
changed, particularly for our legal coefficients. Data available from authors.
336 No fault laws and at-fault people
Methodology
Elizabeth Peters’ study was based on panel data taken from a sample of individuals. By
contrast, our study estimates average filing rates at the state level through state-level time
series cross-sectional data (TSCS). This difference reflects a methodological difference.
Peters’ paper rests on a model in which the crucial building block is the individual. On
social capital theories, however, societies also have their own character, for better or for
worse. We assume that these differences may be observed between states, and we employ
state-wide legal and socioeconomic variables to estimate state divorce filing rates.
The use of TSCS data heightens concerns about idiosyncratic state factors not
captured by the other variables. Because of this, we employed a fixed-effects model
throughout, with a separate intercept for each state.
48
We first estimated divorce levels
48
Judge 1988: § 11.4. On the need to employ a fixed-state effect model for TSCS data, see Gary Becker, (1993).
“Comments on Danzon, Maki, Murray, and Allen.” Journal of Labor Economics New York, NY: MacMillan, 11:S326.
TABLE 5. The determinants of divorce
Variable Name
No-Fault Definition
Marriage, Insurance, and
Working Women
Endogenous
Peters Definition
Marriage, Insurance, and
Working Women
Endogenous
MARRIAGE 0.35010
(7.012)*
PETERS 0.11163
(6.782)*
NO-FAULT 0.098402
(5.925)*
YEAR 0.097344
(4.893)*
0.10388
(5.253)*
EMPLOYMENT GROWTH 0.54357
(5.991)*
0.62611
(6.811)*
UNEMPLOYMENT 0.047327
(2.244)*
0.052766
(2.524)*
METRO 20.037627
(22.051)*
20.028605
(21.620)
INSURANCE 20.61070
(29.844)*
20.57582
(29.434)*
ENTRY 0.037169
(8.306)*
0.037381
(8.666)*
WOMEN IN LABOR FORCE 20.29121
(22.752)*
20.39732
(23.705)*
CATHOLIC 20.11084
(211.19)*
20.12299
(212.41)*
STANDARD ERROR 0.16752 0.16568
R
2
adjusted 0.5935 0.6023
Note: Two-state least squares estimations with fixed state effects. Without economic variables, Peters’
t-statistic is 4.43; No-fault is 2.78.
*significant at .05 level
337BRINIG AND BUCKLEY
through an ordinary least squares (OLS) regression that employed a lagged dependent
variable. However, a Lagrange multiplier test revealed a substantial serial correlation
problem.
49
In theory, one might address this through instrumental variables, but in
practice it is difficult to find instruments that are well correlated with the dependent
variable and are uncorrelated with the errors. Because of this, we employed the Kmenta
cross-sectionally heteroskedastistic and timewise autocorrelated (CHTA) model with
fixed-state effects.
50
The CHTA model assumes cross-sectional independence and b coefficients that do
not vary between cross-sections. However, we adjusted for cross-sectional differences in
two ways. First, we employed a fixed-state effects model, with a separate intercept v
i
for
each State
i
. In addition, the CHTA model corrects for serial correlation through a
state-specific generalized least squares technique. First, the equation is estimated by
OLS. Next, the OLS residuals are used to estimate a separate coefficients of autocor-
relation r
i
(bounded by 21 and 11) for each State
i
. The r
i
’s then are used to transform
the observations to produce a serially independent and homoskedastistic error term.
y
it
5 r
i
X
it
b 1 v
i
1 e
it
Finally, the equation is estimated by the OLS method.
Because several of our variables are substantially endogenous, we also estimated
divorce rates through a two-stage least squares (2SLS) procedure. The decision to
purchase insurance is made often by married couples (particularly if they have chil-
dren) as a form of estate planning, and we would, therefore, expect smaller INSURE
values in states with high divorce rates. We already have noted that marriage rates are
endogenous, insofar as the nonmarried do not divorce. In addition, high divorce rates
may correlate with high marriage rates because of remarriages after divorce.
51
Our
MARRIAGE variable represents married couples per capita. We also employ an endoge-
nous WORKING WOMEN variable, representing the percentage of working age women
with regular outside employment.
52
In a recent paper,
53
Allen Parkman argues that
wives are more likely to be employed in no-fault states (1) to make themselves more
valuable to their husbands to decrease the likelihood of divorce, and (2) to insure
themselves against the financial loss that follows divorce.
Our 2SLS equation, therefore, took the following form:
DIVORCE
it
5 a
0
1 b
1
LAW
it21
1 b
2
UNEMPLOYMENT
it21
1 b
3
EMPLOYMENT GROWTH
it21
1 b
4
YEAR 1 b
5
MARRIAGE
it21
1 b
6
ENTRY
i
1 b
7
METRO
it21
1 b
8
INSURE
it21
1 b
9
CATHOLIC
it21
1 b
10
WORKING WOMEN
it21
1 e
i
,
49
The Lagrange multiplier x
2
for the NO-FAULT Koyck equation was 47.383, and for the PETERS Koyck equation
was 47.080 (both with 23 degrees of freedom). In the non-Koyck equations, the Lagrange multiplier x
2
were 173.857
and 162.311, respectively (again, with 23 degrees of freedom).
50
On CHTA estimation, see Jan Kmenta, (1986). Elements of Econometrics. New York, NY: MacMillan, 618–622.
51
About 75% remarry within 5 years after the divorce. Andrew J. Cherlin, (1981). Marriage, Divorce, Remarriage: Social
Trends in the United State. Cambridge, MA: Harvard University Press, p. 29.
52
Source: Department of Labor, Bureau of Labor Statistics, printouts for various years.
53
Parkman, Allen M. (1998). “Why Are Working Women Working So Hard?” International Review of Law and
Economics 18:41–49.
338 No fault laws and at-fault people
with INSURE, MARRIAGE and WORKING WOMEN taken to be endogenous. As an
exogenous variable, we also employed an inflation-adjusted AFDC variable, represent-
ing the average state payout to a family of four. Because Aid for Families with Depen-
dent Children (AFDC) subsidizes illegitimacy,
54
we expected lower marriage rates in
high AFDC states.
IV. Results
Our results are reported in Tables 4 and 5. Our principal result is that divorce levels are
positively and significantly correlated throughout with no-fault laws, as we have defined
them. Divorce levels also were positively and significantly correlated with the PETERS
no-fault variable in the 2SLS estimations in Table 5 but not in Table 4’s CHTA pooled
regression.
Table 4 reports on a regression of DIVORCE rates on legal, economic, and social
state-level variables. Specifications 1 and 2 employ our NO-FAULT legal variable,
whereas specifications 3 and 4 substitute the PETERS no-fault variable. Because the signs
of our economic coefficients were countercyclical, the economic variables were not
employed in specifications 2 and 4.
Table 5 reports on a 2SLS estimation in which endogenous variables (MARRIAGE,
INSURE, and WORKING WOMEN) are estimated in a simultaneous equation model. The
NO-FAULT variable is employed in specification 1, and the PETERS no-fault variable is
employed in specification 2.
Our NO-FAULT variable is positive and significant throughout. The PETERS variable
also was positive throughout and was significant in the 2SLS equation.
55
This suggests
that the failure of prior studies to detect a significant legal variable might be attributed
to a faulty definition of fault. Alternatively, such studies may have failed to account for
endogenous variables or may have omitted relevant variables.
The economic coefficients are significant throughout. However, the EMPLOYMENT
GROWTH coefficient always has a positive sign, inasmuch as there are more divorces in
expanding economies. The UNEMPLOYMENT coefficient is positive in Table 4, sug-
gesting more divorces in economic downturns, but it is negative in Table 4. These
results suggest that divorce levels might be influenced more closely by social than by
economic variables. Therefore, we dropped both economic variables in specifications 2
and 4 of Table 4.
There were fewer surprises in the social coefficients. The INSURANCE coefficient was
significant and negative throughout, as predicted. A risk-averse society would seem to be
characterized by higher insurance coverage and lower divorce rates, because marriage
serves an insurance function. The METROPOLITAN coefficient was positive in Table 4
but negative in Table 5. A frontier effect continues to be evident in the ENTRY
coefficient, which was positive and significant throughout.
Because divorce rates crested around 1980, at the beginning of our study, we ex-
pected the YEAR coefficient to be negative, as it was in Table 4. However, in the 2SLS
estimation in Table 5, the YEAR and MARRIAGE coefficients were positive. At least part
54
See Margaret F. Brinig and F.H. Buckley, (1998). “The Price of Virtue.” Public Choice (in press).
55
In an OLS model employing a lagged dependent variable, the NO-FAULT coefficient was positive and significant,
whereas the PETERS coefficient was positive and insignificant. In an OLS non-fixed-state effects model, both coeffi-
cients were positive and insignificant when a lagged dependent variable was employed, and they were positive and
significant when it was not. Tables available from authors on request.
339BRINIG AND BUCKLEY
of the decline in divorce rates is attributable to the decline in marriage rates. This also
might explain why the CATHOLIC coefficient was unexpectedly positive and significant
in Table 4, whereas it was negative and significant in Table 5. As expected, WORKING
WOMEN levels seem higher in states with high divorce rates.
V. Conclusion
Our study of divorce rates from 1988 to 1991 provides the strongest evidence to date
that no-fault divorce laws are associated with higher divorce levels. Prior studies failed
to detect a significant no-fault predictor of long-term divorce rates because they defined
“no fault” solely in terms of the dissolution of the marriage and ignored the financial
penalty that a court might impose on an at-fault party. Because of this, they labeled as
no-fault some states that imposed a financial penalty on fault. This is a mistake, because
there will be less fault when it is penalized, and fewer divorces. This explains why our
NO-FAULT coefficient was positively and significantly correlated with higher divorce
rates in Table 4, but why a less precise no-fault predictor was not always significant.
These results are consistent with the theory that the change to no-fault divorce laws
resulted in increased divorce levels. However, our results are suggestive only. In partic-
ular, the assumption that a watertight compartment separates legal and social variables
may be questioned. Divorce levels likely will be lower in societies that stigmatize divorce.
Such societies also are less likely to enact no-fault divorce laws. The legal predictor thus
might serve as a proxy for more fundamental social norms. Although we sought to
address this problem by employing explicit social variables, and by estimating divorce
levels through a 2SLS procedure, we cannot discount the possibility that our NO-FAULT
predictor is, in part, a social variable itself. It is excessive, however, to regard legal
variables, including NO-FAULT as exclusively social. If that were so, then legal rules, and
the economic costs that they impose on rule breakers, would not in themselves affect
behavior. It is as unreasonable to suggest that all causes are social as it is to insist that
social stigma counts for nothing.
This article suggests a solution to the divorce law controversy that might appeal to all
sides in the debate. The consequences of divorce are troubling. Children are harmed
by it, and no-fault laws have impoverished many women. More broadly, the move to
no-fault may have contributed to a coarsening of American society. Nevertheless, many
Americans will resist any attempt to condition the divorce decree on proof of fault. Yet,
if no-fault laws are to be retained at the divorce stage, matrimonial fault still might be
penalized at the stage when assets are divided and spousal support is awarded. Many
states, such as Virginia, have in fact adopted such a regime. Our findings suggest that
divorce rates would be lowered if more states penalized matrimonial fault in this way.
340 No fault laws and at-fault people
