Temporal Differences in Remarriage Timing: Comparing Divorce and Widowhood

Abstract

One gap in the remarriage literature to date concerns the timing of remarriage among different groups. This paper begins to fill this gap by examining the tempo of remarriage among individuals whose first marriages ended in divorce and individuals whose first marriages ended in spousal death. Drawing on event-history models, the results suggest that divorced individuals remarry quicker than individuals whose first marriage ended in spousal death. Interestingly, results also indicate that this relationship is moderated by both gender and parity, suggesting demographic and life course factors can impede or encourage post-marital union formation.

The United States, along with other Western nations, has been witness in recent times to dramatic changes pertaining to union formation patterns. However, large changes in cohabitation, nonmarital childbearing, and union dissolution have been unable to deter most Americans from their desire to marry (Thornton & Young-DeMarco, 2001), regardless of social class or race-ethnicity (Edin & Kefalas, 2005). Consequently, marriage is of considerable importance to the public, social scientists, and policymakers alike. Because marriage is simultaneously a very public institution and the most intimate of all personal relationships (Fomby & Cherlin, 2007), significant attention has been devoted to studying the benefits of marriage, both for individuals (Amato, 2010; Hawkins & Booth, 2005; Johnson & Wu, 2002) and society more generally (McLanahan, Amato, & Furstenberg, 2007).

However, the vast majority of this attention has focused on first or prevailing marriages, (but see De Jong Gierveld, 2004 for an exception using Dutch data; Sweeney, 2010). While first marriage, because of its ubiquity, clearly merits such attention, remarriage has not received as much consideration. However, social and cultural change, such as the rise of expressive individualism and the emphasis on personal fulfillment over the past 50 years, shifts in the economic structure of the labor market, and the redefinition of gender roles and expectations, have contributed to an increase in the number of higher-order marriages (Coontz, 2005; Degler, 1980; Fan, Thompson, & Wang, 1999; Lesthaeghe, 2010; Mintz & Kellogg, 1989). As a result, more than one-fifth of Americans over the age of 40 in 2007 had been married multiple times (United State Census Bureau, 2007). Furthermore, almost half of all marriages in a given year are expected to involve at least one partner with previous marital experience (Ozawa & Yoon, 2002). As these changes continue to take place in the United States, there is little reason to expect any reversal in these trends. Therefore, remarriage patterns are of substantive interest to both policymakers and scholars alike because contemporary rates and patterns of American remarriages have implications at both the micro and macro level. At the micro level, remarriage has been tied to increases in mental and physical health, as well as greater economic stability among the remarried, although substantial gender differences remain (citation blinded for review). At the macro level, the determination of many Americans to remarry following the dissolution of a first marriage can be viewed as evidence that marriage remains a valued social institution, capable of bestowing desired social, emotional, and economic benefits (Sweeney, 2010).

Thus, researchers, policymakers, and practitioners are in need of more information regarding patterns of remarriage in the United States. One gap in the remarriage literature to date concerns the timing of remarriage among different groups. Comparatively little attention has been paid to differences in the timing to remarriage between sub-population. This paper aims to begin to fill this gap by examining the tempo of remarriage among individuals whose first marriages ended in divorce and individuals whose first marriages ended in spousal death. Additionally, we examine whether demographic and life course factors such as parity and gender moderate any associations observed. Given variance in the circumstances under which divorce and spousal death occur and concomitant differences in the social, emotional, and economic support each transition requires, this is an important oversight. To do so, we used data from the National Survey of Families and Households (NSFH), a nationally representative study of adults aged 18 to 95. Our findings suggested important differences between divorcés and widows in the tempo of remarriage, and that gender and parity played a significant role in the timing to remarriage.

Differences in Tempo: Divorce versus Widowhood

Our ability to understand differences in remarriage timing between divorced and widowed individuals depends largely on an expectation that their remarriage patterns vary. Despite the comparative lack of scholarly attention on the tempo of remarriage d (but see de Graaf & Kalmijn, 2003 for an example using Dutch data and Mott & Moore, 1983 for an example using U.S. data), we can draw upon theories generally applied to first marriage to understand why remarital timing may differ between widows and divorcés. These theories fall into two broad categories.

The first theory, marital search theory, suggests that people seeking marriage partners do so within the limits of a marriage market where seekers desire to maximize the quality of the match, but are willing to settle for a minimally acceptable match, called a reservation quality partner (England & Li, 2006; Oppenheimer, 1988). The search occurs in a marriage market characterized by opportunities and constraints imposed on and by both the market and the individual marriage seeker. Market factors influence the availability of marriage partners—especially those who meet or exceed minimally acceptable standards. Individual characteristics, such as educational attainment or parental status, affect one’s ability to maximize spousal quality (Mare, 1991). Thus, this theory suggests that the conditions, preferences, and marriage markets in which the search for a second marriageable partner occurs may be different for divorcés and widows, and these differences could account for any differences in remarriage timing between the two groups.

The second broad category of theories suggests that marital searches can be affected by the unique demographic and life course factors experienced by divorcés and widows. We discuss three possible life course and demographic characteristics here. First, differences in the age at which divorce and widowhood typically occur may lead to varying temporal patterns of remarriage. The median age of divorce was 35 for men and 33 for women, while the median age of widowhood was over 60 for both men and women (United State Census Bureau, 2008). If dissolution occurred at a younger age, the benefits of remarrying, such as another income or greater help with parenting, may outweigh the costs of doing so, such as integrating into and dealing with new kinship relationships (Sweeney, 1997)(Bumpass, Sweet, & Martin 1990; Sweeney, 1997). If dissolution transpired at later ages, the opposite may be true. Second, the exit status, or how the first relationship ended, could account for differences in observed remarriage patterns between divorcés and widows. Divorce, for at least one of the two people, is a voluntary exit. This means that both parties have to acknowledge to themselves that the marriage failed (Sweeney, 2002). In contrast, widowhood is an involuntary exit from marriage (Wu & Schimmele, 2005)—except in rare cases of mariticide or uxoricide. Differences in the ensuing healing and legal processes may also create differences in the tempo of subsequent marriage (Southgate & Roscigno, 2009). Third, differences in the commencement of a new marital search may also explain why divorcés may remarry more quickly than the widowed. Divorcés, because they can foresee the impending disintegration of their marriage, may begin their search for a new partner well before final divorce papers are filed and approved (Amato & Hohmann-Marriott, 2007; Southgate & Roscigno, 2009). For individuals whose marriages ended in spousal death, the search is likely to begin several months, if not years, after marital dissolution (Wu & Schimmele, 2005). Thus, even if both groups spent the same amount of time courting their prospective spouses, we still expect that divorcés would remarry more quickly than widows would because their search began sooner.

Moderating Factors: Gender and Parity

There is also good reason to believe that the relationship between marital exit status and remarriage timing may vary across demographic characteristics, such as gender and parity. Regarding gender, widowed women may have a very low likelihood of remarrying because of the prevailing sex ratio imbalance at later ages. The sex ratio between age 35–39 was approximately 101 men for every 100 women, but dropped to less than 91 men for every 100 women between ages 60 and 64, without accounting for individuals who are married (United State Census Bureau, 2008). Selective mortality by gender means the likelihood that the marriage market for older individuals will be saturated with women is high. Given that hypogamy, or marriages where women are older than their husbands, is non-normative, particularly for unions formed at older ages, it seems likely that whether one’s first marriage ended in divorce or spousal death on remarriage may significantly vary by gender. Older women may experience lower rates of remarriage. Because spousal death tends to occur at older ages, this may produce differences in timing of remarriage between widows and divorcés.

Additionally, the needs of children, the difficulties associated with single parenting, and increased financial strain may combine to increase the likelihood of remarriage among divorced women compared to divorced men and the widowed (Stewart, Manning, & Smock, 2003; Sweeney, 1997). For example, we may observe gender differences in the timing to remarriage because the economic consequences of divorce differ for men and women, where women often experienced a decrease in socioeconomic well-being whereas formerly married men often remained financially stable (Avellar & Smock, 2005), although most men experienced at least some decline as well (McManus & DiPrete, 2001). Thus, because the economic penalty of divorce may be more drastic for divorced women than for divorced men or widowed men or women, divorced women may be more likely to remarry than others may. In addition, women are much more likely to care for children after marital dissolution than are men (Goldscheider & Sassler, 2006), a point we return to below. As a result, we expected variation across both exit status and gender. The results will help clarify the somewhat ambiguous findings in previous literature regarding about differences in the likelihood of remarriage between divorced men and women, where some studies found that men’s odds of remarrying were significantly higher than women’s, while other studies showed a null or even opposite effect (for a full discussion see Coleman, Ivani-Chalian, & Robinson, 2008; Sweeney, 2010).

In terms of parity, the reasons to expect differences in the remarriage rates between divorcés and widows by number of children were relatively straightforward. Remarriage represents an opportunity to introduce another parent into the household who, if chosen wisely, will likely represent both an additional authoritative presence in the household as well as a contributor to the family’s economic well-being (Goldscheider & Sassler, 2006; Sweeney, 1997). Thus, for each additional child, the benefits of remarrying may outweigh the costs of integrating the new partner into the household and kinship structures and dynamics. Some evidence has suggested that parity may increase the likelihood of remarriage for divorced men and women because they partner with individuals who are also parents (Goldscheider & Sassler, 2006). But some research has also suggested that children may represent a detriment to divorced women on the marriage market, whereas, under certain circumstances, being considered a ‘good’ or ‘family-oriented’ father by potential partners may prove beneficial to union formation potential for some men (Stewart, Manning, & Smock, 2003). Furthermore, the average divorcé was more likely to have children living with him or her, whereas the children of widows were more likely to live independently (Wu & Schimmele, 2005) because of the age at which divorce and widowhood typically occur (see above). Many divorcés were in their prime childbearing and childrearing years when their first marriages dissolved and a high percentage of divorced men and women reported wanting an additional child or children (Guzzo & Furstenberg, 2007; Thomson & Li, 2002). Conversely, widowed men and women were less likely to want additional children, which may dampen the desirability of remarriage. Thus, we also expected differences in remarriage rates divorcés and widows by parity.

Data, Method, and Variables

To examine differences in remarriage rates between individuals whose first marriages ended in divorce or spousal death, we used the National Survey of Family Households, a nationally representative sample of over 13,000 individuals collected in 1988. The original sample of 13,000 yielded a final analytical sample of 2,833 individuals whose first marriages had ended at wave 1. Although the data are over 20 years old, they remain the only source for nationally representative data that include explanatory variables for the relationship between divorced/widowed status and remarriage and that contain a sufficiently large age range to compare divorcés and widows. Alternative datasets, such as the National Longitudinal Survey of Youth 1979 (NLSY79), the National Longitudinal Study of Adolescent Health (Add Health), or the Health and Retirement Study (HRS), are all age-restricted, leading to an insufficient number of either widows or divorcés to obtain reliable and unbiased estimates.

Because the fundamental question of this paper involved differences in the timing of remarriage between divorcés and widows, we employed Cox proportional hazards models in the analysis. Cox models, as they are commonly known, simplified the need to specify the functional form of time, which often complicates event history models. To do this, these models made no assumptions about the actual functional form of how the hazard of remarriage, in this case, varies with time. Rather, the Cox model assumed that whatever this functional form was, the hazards of the event occurring remained consistent across time and across levels of the explanatory covariates. Additionally, the Cox model had no intercept, because any value the intercept assumed would simply change the definition of the baseline hazard, which was left undefined as a fundamental part of the model. Each of these points is made clear in the following equation, which provides the model estimated in Table 2, the focus of this paper.

$$h(t_{({\mathit{\text{Remarriage}}}_{\mathit{\text{ij}}})}|{\mathit{X}}_i)=h_0(t)\text{exp }({\beta }_1{x}_{1({\mathit{\text{Divorced}}}_i)}+{\beta }_2{x}_{2(\mathit{\text{Cohabited Prior to }}1{\mathit{\text{st Marriage}}}_i)}+{\beta }_3{x}_{3(\mathit{\text{Duration of }}1{\mathit{\text{st Marriage}}}_i)}+{\beta }_4{x}_{4({\mathit{\text{Female}}}_i)}+{\beta }_5{x}_{5({\mathit{\text{Parity}}}_i)}+{\beta }_6{x}_{6({\mathit{\text{Educational Attainment}}}_i)}+{\beta }_7{x}_{7({\mathit{\text{Working Status}}}_i)}+{\beta }_8{x}_{8({\mathit{\text{Income}}}_i)}+{\beta }_9{x}_{9({\mathit{\text{Catholic}}}_i)}+{\beta }_{10}{x}_{10({\mathit{\text{Baptist}}}_i)}+{\beta }_{11}{x}_{11({\mathit{\text{Not Religious}}}_i)}+{\beta }_{12}{x}_{12({\mathit{\text{Black}}}_i)}+{\beta }_{13}{x}_{13({\mathit{\text{Hispanic}}}_i)}+{\beta }_{14}{x}_{14({\mathit{\text{Other Race}}}_i)}+{\beta }_{15}{x}_{15({\mathit{\text{Age}}}_i)}+{\beta }_{16}{x}_{16(\mathit{\text{Decads of }}1{\mathit{\text{st Marriage}}}_i)}$$ (Equation 1)

Thus, the expected values of the hazard of remarriage (h(t)), conditional on the vector of covariates denoted by X, is defined as a function of the exponentiated log hazard of each coefficient multiplied by the value of the covariate, such as whether one’s first marriage ended in divorce or spousal death.

TABLE 2.

Proportional Hazards Models Predicting the Timing of Remarriage among Individuals Whose First Marriage Ended Either in Divorce or Spousal Death

	Base Model		Gender Interaction		Parity Interaction
	B	SE B	B	SE B	B	SE B
Divorce	0.39***	(0.09)	−0.07	(0.15)	0.09	(0.13)
Prior cohabitation	−0.28***	(0.09)	−0.27**	(0.09)	−0.28**	(0.09)
Duration of 1st marriage	−0.01***	(0.00)	−0.01***	(0.00)	−0.01***	(0.00)
R is female	−0.24***	(0.06)	−0.8***	(0.16)	−0.26***	(0.06)
# of children	0.07***	(0.01)	0.07***	(0.01)	−0.01	(0.03)
R's educational attainment	−0.03	(0.02)	−0.03	(0.02)	−0.03	(0.02)
R worked during first year of marriage	0.03	(0.06)	0.04	(0.06)	0.03	(0.06)
Household income	0.01***	(0.00)	0.01***	(0.00)	0.01***	(0.00)
R is Catholic	−0.24***	(0.07)	−0.24***	(0.07)	−0.23**	(0.07)
R is Baptist	0.01	(0.07)	0.01	(0.07)	0.01	(0.07)
R is not religious	−0.1	(0.10)	−0.1	(0.10)	−0.1	(0.10)
Race-Black	−0.65***	(0.09)	−0.66***	(0.09)	−0.66***	(0.09)
Race-Hispanic	−0.22	(0.13)	−0.21	(0.13)	−0.23	(0.13)
Race-Other	−0.18	(0.30)	−0.14	(0.31)	−0.17	(0.30)
Age	−0.03***	(0.01)	−0.03***	(0.01)	−0.03***	(0.01)
Decade of marriage	−0.46***	(0.06)	−0.47***	(0.06)	−0.47***	(0.06)
Divorce*female			0.62***	(0.17)
Divorce*# of children					0.10**	(0.04)

Note. References are spousal death (Divorce), no prior cohabitation, R is male, R did not work during first year of marriage, other religion, and White.

^*p < .05.

^**p < .01.

^***p < .001.

Variables

The primary variable of interest here was whether an individual’s first marriage ended in divorce (1=divorcé) or spousal death (0=widow). In an attempt to ensure unbiased and efficient estimates of observed differences in remarriage between these two groups, we also included a set of other variables that are associated with both the outcome of remarriage and divorcé/widow status. Additionally, we also included controls for variables known to be associated with the hazard of remarriage.¹ These included whether an individual cohabited with their spouse prior to their first marriage (1=yes), the duration (in months) of their 1^st marriage, gender (1=female), the number of children the respondent has (a continuous measure), the respondent’s educational attainment (0=less than H.S., 1=H.S. degree, 2=Some College, 3=Associates/Vocational Degree, 4=At least Bachelor’s Degree), whether the respondent worked during the first year of their first marriage (1=yes), the respondent’s household income (in $1,000 increments), as well as measures of the respondent’s religious affiliation (Catholic, Baptist, No Religion, or Other Religion, with Other Religion as the reference category), race-ethnicity (Black, Hispanic, Other, and White as the reference category), age (measured continuously), and the decade in which the first marriage began (1900s–1980s). Note that because the data came from a cross-sectional survey, it was not possible to employ time-varying covariates.

Results

To begin, Table 1 presents the mean, standard deviation, and range of all variables included in the analysis. Thus, of the 2,833 respondents whose first marriages had ended (74% of which divorced), almost three-fifths (59%) had remarried, meaning that 41% of cases in our sample are censored in the following models. Our sample was middle-aged, heavily female, somewhat educated, had an average of 2.5 children, and was largely white. Note that the data in this table were weighted, although remaining analyses are not, due to reasons described by Winship and Radbill (1994).

TABLE 1.

Mean, standard deviation, and range for all variables in the analysis

	Mean	S.D.	Min	Max
Remarriage	0.59	0.49	0	1
Divorce	0.74	0.44	0	1
Prior Cohabitation	0.11	0.31	0	1
Duration of 1st marriage (Months)	170.01	163.49	0	782
R is female	0.63	0.48	0	1
# of children	2.51	1.96	0	15
R's educational attainment	1.50	1.31	0	4
R worked during first year of marriage	1.43	0.50	1	2
Household Income	26.06	33.31	0	556.5
R is Catholic	0.21	0.41	0	1
R is Baptist	0.21	0.41	0	1
R is not religious	0.07	0.26	0	1
Race-Black	0.09	0.29	0	1
Race-Hispanic	0.04	0.20	0	1
Race-Other	0.01	0.10	0	1
Age	50.48	15.69	19	87.42
Decade of Marriage	5.37	1.56	1	8

Note: Data in table are weighted to represent estimates of the population means.

Figure 1 displays the observed hazard of remarriage and compares individuals whose first marriages ended in divorce with those whose first marriages ended in spousal death. Based on this figure, it appears that divorcés had a higher risk of remarriage at every point in time than widows do, and Figure 2, which presents the cumulative hazard of remarriage for the same group, confirmed this. From this figure, we can see that the estimated hazard of remarriage was higher for divorcés than widows at every point in time.

FIGURE 1.

Observed Hazard of Remarriage among Divorcees and Widow(er)s in months after first marital dissolution.

FIGURE 2.

Observed Cumulative Hazard among Divorcees and Widow(er)s in Months after First Marital Dissolution.

However, we also use statistical models that control for confounding factors in order to understand if divorcés or widows remarry quicker. To do this, we estimated the model described in Equation 1. The results are found in Table 2.

Here, we found that the hazard of remarriage was nearly 50% higher (e^.39=1.48) for divorcés relative to widows, indicating that being divorced was associated with an increased hazard of remarriage, thereby speeding up the tempo of reentry into marriage. This is in line with our expectations based on the Figures 1 and 2. In terms of the other covariates, those who cohabited prior to their first marriage experienced a lower hazard, as did females, Catholics, Blacks, individuals whose first marriages lasted longer and who were married in recent decades, and older respondents. In contrast, increasing parity and income were each associated with a greater hazard of remarriage. Thus, these results largely supported our expectations regarding differences in remarriage rates.

However, there was also reason to believe that the tempo of remarriage for divorcés and widows may vary by gender and parity. To examine this possibility, we included interactions between the divorce variable and the gender and parity variables, respectively. The results are found in the second and third column of table 2 as well as Figures 3 and 4. For gender, we found a significant interaction between the two variables, confirming our intimations. The effect of being divorced compared to widowed on the hazard rate of remarriage was 86% (e^.62=1.86) higher for females than it is for males. Thus, although females experienced an overall lower hazard of remarriage, net of all other variables included in the model, this was driven largely by the low hazard of remarriage for females respondents whose first marriages ended in spousal death. As can be seen in Figure 3, female divorcées had the highest hazard of remarriage, while widows had the lowest. The difference in remarriage hazards between male divorcés and widowers was smaller, although the remarriage advantage for divorcés remained, with divorcés reporting a higher hazard of remarriage than widowers do.

Figure 3.

The predicted hazard of remarriage among widowed and divorced individuals by gender

Figure 4.

The predicted hazard of remarriage among widowed and divorced individuals, by # of children

Figure 4 graphs the interaction between parity and divorcé/widow status, based on estimates obtained from Table 3 (Figure 3 was obtained similarly). Here, the differences in the effect of divorced/widowed status on remarriage as parity changes were quite stark. For the spousal death group, parity did not appear to be linked in any meaningful way to the hazard of remarrying, remaining comparatively flat as parity rises. In contrast, increasing parity did appear to be linked to higher hazards of remarrying among respondents whose first marriage ended in divorce. Figure 4 demonstrated the growing hazard for this group with the addition of each child. Compared to divorced respondents with no children, the hazard of remarriage was 52% greater (e^.10=1.11=1.11^4=1.52) for those with three children. Note that although the estimated hazard rates on the y-axis of these figures may appear small, these represent the (instantaneous) hazard at any given point in time, so the difference in the overall, cumulative hazard between divorcés and widows are similar to those observed in Figure 2.

TABLE 3.

Model Fit and Deviance-Based Hypothesis Tests

	Null	Base Model	Divorce* Female	Divorce* Children
Log Likelihood	−13634.3	−10835.4	−10829.3	−10830.9
# of Parameters	0	16	17	17
Deviance	27268.6	21670.8	21658.5	21661.7
AIC	27268.63	21702.78	21692.53	21695.72
BIC	27268.63	21797.96	21793.67	21796.86
Deviance-based Hypothesis Tests
H0:βx=0		0.00***	0.00***	0.00***

Table 3 displays model fit and deviance-based hypothesis tests. The log likelihood, AIC, and BIC all indicated increased model fit with the inclusion of new variables into the model. To examine this in greater depth, we performed deviance-based hypothesis tests, which confirmed that the inclusion of each new set of variables significantly improved model fit. These results all suggest that our models fit the observed data well, lending further support to the conclusion that individuals whose first marriages ended in divorce tend to remarry more quickly than individuals whose first marriages ended in spousal death.

Discussion

The goal of this paper was to examine differences in the timing and tempo of remarriage among individuals whose first marriages had ended either in divorce or spousal death. Results suggested that divorced individuals tended to remarry quicker than individuals whose first marriage ended in spousal death, with divorced individuals experiencing a higher hazard of remarriage at every point in time than widowed respondents. Interestingly, results also suggested that this relationship was moderated by gender and parity.

This research provided empirical, quantitative evidence that divorced individuals tend to remarry quicker than did widowed individuals. It is likely that this difference is explained by variation in the conditions, preferences, and the market conditions in which the search for a new partner occurs, as marital search theory would predict. For example, because widowhood generally arises at later ages, difficulties of incorporating a new partner into existing familial and kinship networks may outweigh the benefits a new partner bestows. The fact that divorce is often a voluntary exit for individuals may also lead people to begin seeking and marrying a new partner sooner because divorced individuals seek to reestablish their identity. In contrast, the involuntary split from their partner may take much longer to recover from, leading to delays in remarriage, relative to their divorced counterparts, not least because divorced individuals likely begin their search for a new marital partner before the divorce is official (Amato & Hohmann-Marriott, 2007; Southgate & Roscigno, 2009).

Interestingly, we also found that these differences observed between divorced and widowed individuals in the tempo of remarriage differed by gender and parity. The hazard of remarriage was highest for divorced females, while it was the lowest for their widowed counterparts. Given the imbalanced sex ratio at older ages, with the marriage market at older ages being saturated with older women, this is not surprising, especially in light of the non-normative nature of hypogamy for women.

In terms of parity, the number of children one has appears to have little to no effect on the hazard of remarriage among the widowed, whereas parity had a strong and consistently positive effect among the divorced. Because children may represent both an obstacle to union formation (Stewart, Manning, & Smock, 2003), the explanation for this lies largely in the age at which divorce and widowhood typically occur. Children of widowed individuals are typically older, meaning that they are no longer living with their parents. In contrast, divorced individuals are often young enough to have children still living at home. Thus, concerns about childbearing will likely impinge more on remarriage decisions for divorced than for widowed individuals, and this is precisely what we found.

However, these findings are not without limitations, the most pivotal of which is the potential for endogeneity regarding certain variables shown to be associated with the hazard of remarriage, such as income. For instance, because income was measured at the time of the survey rather than the time of the observed marital transitions, the possibility of reverse causation cannot be ignored. Our measure of income could be the result of either the remarriage or the initial divorce, and scholarly work has demonstrated important differences in socioeconomic status after union dissolution (Avellar & Smock, 2005). While we feel that the intransigence of social class relations across the life course gives us reasonable assurance that our measure of income yields an admittedly crude measure of financial resources available to the respondents, future work would do well to work toward a more refined model of remarital timing by parsing out the time-ordering issues that we cannot address here due to data limitations. Furthermore, this paper has focused solely on second marriages, thereby neglecting higher-order marriages. Because of the constant flux in romantic relationships (Fomby & Cherlin, 2007), the number of third and higher-order marriages has increased, and selectivity into these higher order marriages may be partly reflected in the results estimated here. The fact that we lack information on cohabitation also constitutes a deficiency of the study, especially if premarital cohabitation has a negative influence on subsequent marital outcomes (Thornton, Axinn, & Hill, 1992). Unobserved heterogeneity from other confounders remains an issue, particularly as related to relationship skills and who initiated the divorce. The fact that we cannot tell who initiated the divorce nor anything about the quality of the former spouses’ interactions means that these models may be biased to the extent that such variables are not correlated with any variables in our model. Additionally, the fact that we had to rely on theories designed primarily to explain patterns in first marriages points to a severe need in the literature for theoretical development. This is likely a fruitful area for qualitative work. By interviewing people whose first marriages have ended and employing a grounded theory approach, qualitative work could shed much needed theoretical light that could help explain differences in the tempo to remarriage for certain groups of individuals.

This paper contributes to the literature in several important ways. First, this paper adds to a burgeoning literature on post-marital union formation by describing temporal differences in the hazard rate of remarriage between individuals whose first marriages ended in divorce or spousal death, and found that divorced individuals were more likely to remarry at every point in time than their widowed counterparts were. Second, we provide evidence that demographic variables such as gender and parity moderate this association. Future work should attempt to better understand the mechanisms underlying and explaining this moderation. Doing so will allow us a better understanding of the nuanced and complex pathways between marital dissolution and subsequent union formation.