Does the Rebound Effect Exist?

Abstract

Self-help books often advise readers to avoid rushing into new relationships after a break-up. To date, there is little evidence supporting this recommendation. This paper tests the effects of rebound time, measured as time elapsed between marital dissolution and the formation of a new union, on remarriage duration. Data from the first wave of the National Survey of Families and Households and generalized additive models reveal no evidence of a rebound effect. This finding holds after adjusting for various demographic differences between respondents.

Full text

Does the Rebound Effect Exist? Time to Remarriage and Subsequent Union Stability
Nicholas H. Wolfinger
Department of Family and Consumer Studies, University of Utah
ABSTRACT
Self-help books often advise readers to avoid rushing into new relationships after a
break-up. To date, there has been little evidence supporting this recommendation. This
paper tests the effects of rebound time, measured as time elapsed between marital
dissolution and the formation of a new union, on remarriage duration. Data from the
first wave of the National Survey of Families and Households and generalized additive
models reveal no evidence of a rebound effect. This finding holds after adjusting for
various demographic differences between respondents.
Running head: Does the Rebound Effect Exist?
Published in Journal of Divorce & Remarriage 46 (2007): 9-20

Does the Rebound Effect Exist?
1
INTRODUCTION
Many self-help books offer a piece of advice that is simple and familiar: after a
relationship ends, wait before starting a new one. The following passages are typical:
"Rebound" relationships can also pose a problem. Some persons, after the
breakup of a long-standing relationship, are tempted to rush into a new
relationship. . . . They commit to a serious relationship prematurely. Their
"rebound" union will work for awhile, but in time overlooked and unresolved
difficulties will have to be faced and settled (Snyder, 1993, p. 122).
You can avoid the pain of a rebound relationship by following one simple rule:
stay away from new relationships until you know you are good and ready. . . .
There's no harm in rebound relationships if you take them for what they are—
temporary and not particularly meaningful (Forrester, 2005).
This paper analyzes National Survey of Families and Households data to determine
whether rebound time, defined as months elapsed between an initial divorce and
subsequent remarriage, affects that remarriage's stability. The issue is noteworthy
given that over two-thirds of divorcées remarry (Bramlett and Mosher 2001; Kreider and
Fields 2001).
No studies have tested the rebound hypothesis since Aguirre and Parr in 1982.
Analyzing data from the National Survey of Family Growth, they found no evidence of a
rebound effect. However, above and beyond its antiquity their study suffers from three
shortcomings: 1) They only examine time to remarriage, failing to take premarital
cohabitation into account. Cohabitation is especially common among divorced people;
by extension, many individuals in second marriages first live together (Bumpass & Lu,

Does the Rebound Effect Exist?
2
2000). 2) Aguirre and Parr (1982) did not employ event history methods (e.g., Allison,
1995), nowadays considered indispensable for studying marital stability. 3) Potential
rebound effects were constrained to linearity. As I suggest in the follow paragraphs,
some plausible explanations for rebound effects imply nonlinear relationships to divorce
risk. The current study overcomes all of these limitations.
The self-help and research literatures offer several reasons why hasty
remarriage might be correlated with higher rates of divorce (there are no arguments to
suggest lower dissolution rates for second marriages quickly following first divorces).
These will now be considered. Two of these arguments suggest specific functional
forms in the relationship between rebound time and divorce risk, so the data may allow
me to adjudicate between them.
Inadequate Search Times
Remarrying quickly may represent an inadequate search process: overly eager
divorcées may choose second husbands or wives who are not good bets for lasting
unions. But research by South (1995) casts doubt on this hypothesis. He shows that
marriage market characteristics like the availability of marriage-appropriate singles
cannot account for the well-known relationship between youthful marriage and divorce.
This finding suggests that premature marriage does not reflect the failure to consider
spousal alternatives, potential mates that might otherwise provide inducements to
divorce. Since South analyzed National Longitudinal Survey of Youth data, his
respondents were young and therefore mainly in first marriages. Accordingly, his
findings may not hold for higher order unions.
Adultery, Selection, and Divorce Proneness

Does the Rebound Effect Exist?
3
It has been suggested that second marriages have high divorce rates because they
draw on a population that has already demonstrated its willingness to dissolve
marriages (Martin & Bumpass, 1989). Put another way, people in second marriages are
divorce prone. Why is this the case? Take the example of people who married the
adulterous partners involved in the break-up of their initial marriages. This would
naturally produce short rebound times. These marriages are comprised of people who
have already demonstrated their willingness to commit adultery, so they may do so in
their second marriages as well. Although it is not known how many second marriages
involve adulterous partners, extramarital affairs may occur in over 40% of dissolved
marriages (South & Lloyd, 1995; Stewart et al., 1997). Perhaps a meaningful number of
remarriages therefore result from adultery in first marriages. In turn, the willingness to
philander may ultimately sabotage second marriages. If this line of reasoning holds, it
implies a selection mechanism to account for the relationship between short rebound
times and high divorce rates in remarriages. This would cause the likelihood of divorce
to spike at short rebound times, reflecting adulterously divorce-prone individuals who
rush into second marriages, then decrease monotonically.
Inadequate Postdivorce Recovery
Many of the major real-world transitions associated with marital disruption, such as
residential mobility, take place within a year or two (McLanahan, 1983). On the other
hand, the clinical literature does not offer precise insight into how long the emotional
recovery from divorce takes. The best-known account, from Judith Wallerstein and her
colleagues, is typical in its vagueness:
The average woman was well into her third post-separation year before life
assumed a new coherence and stability; the average man accomplished this

Does the Rebound Effect Exist?
4
restabilization earlier, within the second year. . . . One sobering finding was that
almost five years post-separation, 31 percent of the men and 42 percent of the
women had not yet achieved psychological or social stability (Wallerstein & Kelly,
1980, pp. 190, 191).
It can more reasonably be claimed that different components of emotional well-being
return at different rates (Stewart et al., 1997; Wallerstein & Kelly, 1980). The upshot is
a theme expressed by nearly all self-help books addressing rebound relationships: "A
new relationship cannot begin until you have grieved the last relationship (Mellody &
Freundlich, 2003, p. 139)." Therefore, a longer wait after marital disruption may
produce greater emotional health, and potentially more stable remarriages. But how
long? The ambiguity and conflicting results characterizing the extant literature makes it
difficult to offer a precise estimate. It is safer to say that different people recover at
different rates, so the likelihood of divorce should decrease gradually and monotonically
as rebound time increases.
METHODS
Data
This research uses data from the National Survey of Families and Households
(NSFH), a national sample survey of American adults 19 and over (Sweet, Bumpass, &
Call, 1988). Thirteen thousand and seven respondents were interviewed in 1987 and
1988. These include a main sample of 9,643 plus an oversample of minorities,
newlyweds, single parents, individual parents in stepparent families, and individuals in
cohabiting unions. Although many NSFH respondents participated in two follow-up
interviews, only the Wave 1 data are analyzed. Prospective data from two or three

Does the Rebound Effect Exist?
5
waves offer no great advantage and would yield a prohibitively small sample.
Analysis is limited to female respondents who remarried after an initial divorce (N
= 1,171). Men (N = 740) are omitted for two reasons. First, men's marital histories
have long been known to be comparably unreliable (Bumpass, Martin, & Sweet, 1991).
This appears to be the case here, given that the NSFH data contain far fewer remarried
men despite the fact that they remarry at a higher rate than do women. Second,
analyzing only women makes it probable that respondents are living with any minor
children they have adopted or given birth to (90% of divorced mothers have at least
partial physical custody [Cancian & Meyer, 1998]). This is important due to data
limitations discussed below.
NSFH case weights are used so the data comprise a nationally representative
sample. A common strategy for weighted data is to include the variables used to
calculate the weights as independent variables in regression analyses (Winship &
Radbill, 1994), but this is not feasible given the complexity of the NSFH weighting
scheme. Unfortunately sample weights can adversely affect standard errors, resulting
in artificially inflated test statistics. One solution is to estimate Huber-White standard
errors, but these are not available for the generalized additive models employed in this
paper. Accordingly, the computed t-ratios should be viewed as biased upwards.
I delete listwise any cases with missing or invalid data on when respondents' first
marriages ended (N = 57), event histories of second marriages (N = 28), and
cohabitation (N = 13). The paucity of missing data on the independent variables
minimizes the potential utility of strategies like imputation.
Variables
The dependent variable for this paper is second marriage duration. Marriage
start times are measured in two ways: 1) The month of legal marriage; 2) The month

Does the Rebound Effect Exist?
6
spouses began living together, either by marriage or cohabitation. Second marriages
ending in spousal death or intact at the time of the interview are considered censored.
Couples that have been separated for a year or more from their second spouses are
treated as divorced; Bumpass, Martin, & Sweet (1991) show that the chance of
reconciliation at this point is slight. Data on marriage duration, measured in months, are
derived from retrospective marital histories. Summary statistics for this and other
variables are shown in Table 1.
Table 1 Here
I choose not to analyze cohabiting relationships that do not end in marriage.
Cohabiting unions are notoriously unstable, with very high dissolution rates (Bumpass
& Lu, 2000). Perhaps as a result, almost no sociodemographic characteristics affect
union stability. Even age at union formation (Ruf & Qian, 1999) and parental divorce
(Wolfinger, 2001, 2005), two of the strongest predictors of marital stability, make no
difference. Taken together, these findings suggest that attempting to predict
cohabitation stability is not a tenable proposition.
The primary independent variable is rebound time, the number of months
elapsed between the end of respondents' first marriages and when new unions are
formed. Marriage end time is measured when respondents stopped living together; new
union formation is alternately measured as the beginning of their remarriages or the
point at which respondents begin cohabiting with their remarriage partners. All analyses
include one control variable, the century month respondents' first marriages ended in
separation. This is vital given secular trends in both remarriage (Martinson, 1994) and
divorce (Cherlin, 1992).
Additional variables are added to rule out the possibility that any observed effects
of rebound time on divorce risk are the product of sociodemographic differences
between respondents. With one exception, each of these items is correlated both with
time to remarriage and subsequent marital stability. Although the list does not contain

Does the Rebound Effect Exist?
7
every known predictor of divorce, it accounts for most of the important differences that
might produce spurious correlations between rebound time and the stability of second
marriages. Little additional information about former second spouses is available in the
NSFH.
Ethnicity, related both to remarriage and divorce (Bramlett & Mosher, 2001), is
coded as white, Black, and other; white is the reference category. Sample size
considerations preclude more than three categories. Parental divorce decreases the
likelihood of remarriage (McLanahan & Bumpass, 1988) and increases the likelihood
that second marriages will dissolve (Amato & Booth, 1991; Wolfinger, 2000, 2005), so it
is measured with a dummy variable. Older divorcées have both lower remarriage rates
and lower divorce rates in remarriages (Bramlett & Mosher, 2001); age at the end of first
marriage is highly correlated with first marriage length (r = .91), so any argument for
using one in lieu of the other is moot. Children decrease the likelihood of remarriage
(Martinson, 1994) and increase divorce rates in second marriages (White & Booth,
1985), so the presence of minor children at the end of respondents' initial marriages is
measured with a dummy variable. This is constructed with information on fertility,
adoption timing, and childhood mortality; unfortunately, it is not possible to verify that
children are actually living with respondents. However, this is likely given high levels of
female custody. The final independent variable used is education, measured at the time
respondents end their initial marriages. Unfortunately the data are not of adequate
quality to treat education as a time-varying covariate. Although education does not
appear to be related to women's remarriage rates (Martinson, 1994), educated women
report lower divorce rates in second marriages (Bramlett & Mosher, 2001). More
generally, education is such a broad marker of social well-being that it should be taken
into account.

Does the Rebound Effect Exist?
8
Analysis
Rebound time may well have nonlinear effects on the stability of second
marriages. The explanations considered here suggest that the likelihood of divorce
probably decreases as rebound time increases, but it is not known whether this decline
is especially precipitous in the first months after marital dissolution. Perhaps down the
road the likelihood of divorce increases, as the pool of still-single divorcées shrinks to
include only those ill-equipped to succeed at remarriage. These and other mechanisms
may produce otherwise unobserved heterogeneity among remarried people. With this
in mind, there is little way of knowing ahead of time what the functional form of the
relationship between rebound time and remarriage stability will resemble.
The solution is a generalized additive model (Hastie & Tibshirani, 1990; for
overviews see Hastie, 1993; Beck & Jackman, 1998), hereafter referred to as a GAM.
The relationship of each independent variable in a GAM can be specified to have a
traditional linear relationship to the dependent variable, as in the generalized linear
model, or a nonparametric, potentially nonlinear relationship. I use the implementation
developed by Royston and Ambler (2002), where nonparametric relationships are
optimized via cubic smoothing splines; lowess local regression is the alternative. The
effect of each independent variable, whether linear or nonparametric, is net of all other
independent variables as for the generalized linear model. For nonparametric terms, a
likelihood ratio test determines whether the fitted relationship to the dependent variable
departs significantly from linearity. Since marriage duration is a time-dependent
phenomenon, event history analysis is appropriate. I therefore estimate GAMs where
the link function is the Cox proportional hazard model. This accounts for differential
exposure to the risk of marital dissolution, as well as right censoring.
The three temporal predictors in my analyses, rebound time, century month of

Does the Rebound Effect Exist?
9
initial separation, and age at separation, are all treated as nonparametric. Estimates for
these terms are based on four degrees of freedom; preliminary analyses employing
between three and five degrees of freedom did not produce substantially different
results. All other independent variables are categorical and are therefore entered into
analyses as linear predictors.
RESULTS
Table 2 shows the effects of rebound time and other factors on the stability of
second marriages, excluding premarital cohabitation. Model 1 shows that rebound time
has no relationship to divorce risk. The linear coefficient, half the size of its standard
error, is nonsignificant; so too is the likelihood ratio test measuring departure from
linearity (The nonparametric results of GAMs are sometimes shown as plots; I do not do
so since rebound time has no relationship, linear or otherwise, to divorce risk.) In
contrast, the century month of respondents’ initial separation has a statistically
significant and nonlinear association with marital stability. This result is predictable
given that divorce rates rose throughout most of the twentieth century, and particularly
rapidly during the 1965-1979 boom (Cherlin, 1992).
Table 2 Here
Model 2 adds other independent variables, including age at initial separation,
history of parental divorce, presence of children, race, and education. Rebound time
continues to have no effect on divorce rates. The other independent variables have
effects generally consistent with the divorce literature. Older respondents have more
successful second marriages; note also that the relationship of this variable to divorce
risk does not depart significantly from linearity. On the other hand, people from
divorced families and people who did not graduate from high school have higher divorce
rates. African-Americans have higher divorce rates than do members of other

Does the Rebound Effect Exist?
10
population groups. Finally, children make divorce more likely, given the difficulties
inherent to stepfamilies. The predictable effects of these independent variables on
remarriage stability bolster confidence that the data are reliable, which in turn provides
support for the absence of a rebound effect.
Model 3 extends the question of rebound effects to union duration measured by
either marriage itself or the start of premarital cohabitation; Model 4 adds all control
variables to Model 3. As for Models 1 and 2, there is no evidence of a rebound effect.
In other respects the results are similar. Respondents from divorced families,
respondents with children, older respondents, less educated respondents, and
respondents ending their first marriages more recently all have higher divorce rates.
CONCLUSION
This paper has a single straightforward finding: there is no rebound effect.
People quickly entering new relationships after an initial divorce, whether by remarriage
or cohabitation followed by remarriage, do not have higher divorce rates. This finding
persists after controlling for key demographic differences between respondents. The
advice offered by many self-help books ("Don't get into a new relationship too quickly!")
therefore has no basis in reality.
I have examined the rebound hypothesis only as it pertains to marriage. Perhaps
it does hold for dating relationships. Future research might explore this issue should
adequate data become available. The rebound effect may also exist within remarriage
only for certain kinds of people, so analyses employing detailed psychometric measures
might prove fruitful.

Does the Rebound Effect Exist?
11
AUTHOR’S NOTE: Correspondence to Wolfinger, Department of Family and Consumer
Studies, 225 South 1400 East, AEB 228, University of Utah, Salt Lake City, UT 84112-
0080; e-mail: Nick.Wolfinger@fcs.utah.edu. I thank Lisa Diamond, Lori Kowaleski-
Jones, Heather Melton, and W. Bradford Wilcox for useful suggestions, Melissa Dean
for her initial work on this project, and Sonja Anderson for research assistance. The
National Survey of Families and Households was funded by a grant (HD21009) from the
Center of Population Research of the National Institute of Child Health and Human
Development. The survey was designed and carried out at the Center for Demography
and Ecology at the University of Wisconsin-Madison under the direction of Larry
Bumpass and James Sweet. The field work was done by the Institute for Survey
Research at Temple University. An earlier version of this paper was presented at the
2006 annual meeting of the Population Association of America, Los Angeles.

Does the Rebound Effect Exist?
12
REFERENCES
Aguirre, B. E., & Parr, W. C. (1982). Husbands’ marriage order and the stability of first
and second marriages of white and black women. Journal of Marriage and the
Family, 44, 605-620.
Allison, P. D. (1995). Survival Analysis using the SAS System: A Practical Guide. Cary,
NC: SAS Institute, Inc.
Amato, P. R., & Booth, A. (1991). Consequences of parental divorce and marital
unhappiness for adult well-being. Social Forces, 69, 905-914.
Beck, N., & Jackman, S. (1998). Beyond linearity by default: Generalized additive
models. American Journal of Political Science, 42, 596-627.
Bramlett, M. D., & Mosher, W. D. (2001). First Marriage Dissolution, Divorce, and
Remarriage: United States. Advance data from vital and health statistics; no.
323. Hyattsville, MD: National Center for Health Statistics.
________. (2002). Cohabitation, Marriage, Divorce, and Remarriage in the United
States. National Center for Health Statistics, Series 23, Number 22.
Bumpass, L. L. & Lu, H.-H. (2000). Trends in cohabitation and implications for children's
family contexts in the United States. Population Studies, 54, 29-41.
Bumpass, L. L., Martin, T. C., & Sweet, J. A. (1991). The impact of family background
and early marital factors on marital disruption. Journal of Family Issues, 12, 22-
42.
Cancian, M. & Meyer, D. R. (1998). Who gets custody? Demography, 35, 147-157.
Cherlin, A. J. (1992). Marriage, Divorce, Remarriage, revised edition. Cambridge, MA:

Does the Rebound Effect Exist?
13
Harvard University Press.
Forrster, P. (2005). On the rebound: Remember what your mama told you: You can't
hurry love. Rebound Relationships <http://rebound-
relationships.com/rebound.html>.
Hastie, T. J. (1993). Generalized additive models. Pp. 249-308 in Statistical Models in
S, edited by J. M. Chambers & T. J. Hastie. Pacific Grove, CA: Wadsworth &
Brooks.
Hastie, T. J. & Tibshirani, R. J. (1990). Generalized Additive Models. London: Chapman
& Hall.
Kreider, R. M. & Fields, J. M. (2001). Number, Timing, and Duration of Marriages and
Divorces: Fall 1996. Current Population Reports, P70-80. Washington, DC: U.S.
Census Bureau.
Martin, T. C., & Bumpass, L. L. (1989). Recent trends in marital disruption.
Demography, 26, 37-51.
Martinson, B. C. (1994). Postmarital union formation: Trends and determinants of the
competing roles of remarriage and nonmarital cohabitation among women in the
United States. Unpublished doctoral dissertation, Department of Sociology,
University of Wisconsin-Madison.
McLanahan, S. S. (1983). Family structure and stress: A longitudinal comparison of two-
parent and female-headed families. Journal of Marriage and the Family, 45, 347-
357.
McLanahan, S. S., & Bumpass, L. L. (1988). Intergenerational consequences of family
disruption. American Journal of Sociology, 94, 130-152.

Does the Rebound Effect Exist?
14
Mellody, P. & Freundlich, L. S. (2003). The Intimacy Factor: The Ground Rules for
Overcoming the Obstacles to Truth, Respect, and Lasting Love. San Francisco:
Harper Collins.
Royston, P. & Ambler, G. (2002). GAM: STATA module for generalized additive models.
Statistical Software Components S428701, Department of Economics, Boston
College.
Ruf, S. D. & Qian, Z. (1999). Stability of cohabitation: The impact of socio-demographic
characteristics. Paper presented at the annual meeting of the Population
Association of America, New York.
Snyder, J. J. (1993). Marrying for Life: The Challenge of Creating a Lasting Friendship.
Saratoga, CA: R&E Publishers.
South, S. J. (1995). Do you need to shop around? Age at marriage, spousal
alternatives, and marital dissolution. Journal of Family Issues, 16, 432-449.
South, S. J. & Lloyd, K. M. (1995). Spousal alternatives and marital dissolution.
American Sociological Review, 60, 21-35.
Stewart, A. J., Copeland, A. P., Chester, N. L., Malley, J. E., & Barenbaum, N. B.
(1997). Separating Together: How Divorce Transforms Families. NY: The
Guilford Press.
Sweet, J. A., Bumpass, L. L., & Call, V. R. A. (1988). The design and content of the
national survey of families and households. NSFH Working Paper No. 1, Center
for Demography and Ecology, University of Wisconsin.
Wallerstein, J. S., & Kelly, J. B. (1980). Surviving the Breakup: How Children and
Parents Cope with Divorce. New York: Basic Books.

Does the Rebound Effect Exist?
15
White, L. K. & Booth, A. (1985). The quality and stability of remarriages: The role of
stepchildren. American Sociological Review, 50, 689-698.
Winship, C. & Radbill, L. (1994). Sampling weights and regression analysis.
Sociological Methodology and Research, 23, 230-257.
Wolfinger, N. H. (2000). Beyond the intergenerational transmission of divorce: Do
people replicate the patterns of marital instability they grew up with? Journal of
Family Issues, 21,1061-1086.
________. (2001). The effects of family structure of origin on offspring cohabitation
duration. Sociological Inquiry, 71, 293-313.
________. (2005). Understanding the Divorce Cycle: The Children of Divorce in Their
Own Marriages. New York: Cambridge University Press.

Table 1. Percentages or means for variables.
Length of second marriage in months 123 (123)
Second marriage ended in divorce
No 68%
Yes 32
Rebound: Time to remarriage (months) 60 (58)
Rebound: Time to cohabitation (months) 52 (55)
Century month first marriage ended 822 (152)
Age at end of first marriage in months 303 (80)
Children at end of first marriage
No 32%
Yes 68
Grew up in divorced family
No 15%
Yes 85
Education
Less than H. S. 32%
High school graduate 34
Some college 23
College graduate 10
Race
White 79%
Black 15
Other 6
Notes : Numbers in parentheses are standard deviations
Percentages may not sum to 100 due to rounding error

Table 2. Generalized Additive Model Estimates of Remarriage Stability on Rebound Time and Other Factors.
Marriage and
Marriage Only Cohabitation
Linear Departure Linear Departure Linear Departure Linear Departure
Variables Linear from linearity estimate from linearity estimate from linearity estimate from linearity
Rebound time .0005 n.s. .0005 n.s. .0013 n.s. .0014 n.s.
(.0010) (.0010) (.0011) (.0011)
Century month of initial separation .0027*** ** .0032*** * .0026*** ** .0031*** *
(.0004) (.0010) (.0004) (.0004)
Age at initial separation -- -- -.0027*** n.s. -- -- -.0027*** n.s.
(.0006) (.0006)
Respondent from divorced family -- -- .2609* -- -- -- .2467* --
(.1255) (.1255)
Education
Less than H.S. -- -- .2349* -- -- --
.1891
+
--
(.1134) (.1134)
H.S. graduate -- -- -- -- -- -- -- --
Junior college graduate -- -- .0145 -- -- -- .0194 --
(.1302) (.1301)
College graduate -- -- .0912 -- -- -- .0961 --
(.1728) (.1728)
Race
White -- -- -- -- -- -- -- --
Black -- --
.2444
+
-- -- -- .1848 --
(.1267) (.1256)
Other -- -- .0954 -- -- -- .0241 --
(.1828) (.1833)
Children at initial separation -- -- .2431* -- -- -- .2408* --
(.1042) (.1044)
Log likelihood -3034.89 -3011.88 -3048.5 -3029.45
Note
s
: N for all models is 1,171; numbers in parentheses are standard erro
r
+
p < .10; *p < .05; ** p < .01; ***p < .001
Model 1 Model 2 Model 3 Model 4