RELATIONSHIP CONTEXT AND INTIMATE PARTNER VIOLENCE FROM ADOLESCENCE TO YOUNG ADULTHOOD

Recent evidence shows that rates of intimate partner violence peak during young adulthood, with one-third (32%) of young adults reporting victimization and one-quarter (24%) perpetration [1, 2]. Prior studies focusing on IPV during adolescence and young adulthood have been limited to victimization [1] or examined shorter time periods [3], and have not focused sufficient attention on proximal relationship-specific factors as influences on IPV experiences. Drawing on a life course perspective we examine the degree of continuity in IPV (perpetration and victimization), and assess whether individual changes in relationship context, including relationship quality (infidelity, relationship churning, frequency of disagreements trust, validation, self-disclosure, and commitment), relationship type (dating, cohabitation, marriage), and partner continuity (retaining a partner between data points), are associated with changes in IPV across adolescence and young adulthood.

Entry into the world of romantic relationships begins during adolescence. As youth transition to adulthood, romantic relationships increase in duration, take on greater meaning [4], and often follow a pattern characterized by increasing levels of trust, intimacy, and commitment [5, 6]. While these changes largely reflect positive development for most young people [7], considerable variation exists, and factors such as poverty and the accumulation of relationship experiences may complicate the idea of a smooth transition and set of relationship progressions [8]. Consequently, for some individuals romantic relationships may be associated with additional discord, and outcomes such as IPV [9].

A key insight of the perspective is that even though early experiences have a formative influence, all phases of development present new opportunities and challenges. Thus, research has shown that peer relationships are linked to a variety of developmental outcomes, even after controlling for family factors [10]. Even though it is well recognized that over time romantic involvement becomes an increasingly important preoccupation and context for development [7], the impact of romantic relationships in general and in relation to specific outcomes such as IPV has received considerably less attention relative to family and peer factors [6]. Thus, it is important to explore how romantic involvement influences adolescents and young adults, particularly in connection with IPV, as this is a behavior that by definition takes place within the context of intimate relationships.

As relationships increase in salience, they also increase in duration [6], and this combination may result in the perception of a ‘higher stakes’ relationship, and a longer window for the development of conflicts–some of which may escalate to the point of violence. Furthermore, continued IPV within a relationship is consistent with the idea that some couples develop interaction styles that once developed, may be intractable or difficult to extinguish [3].

During the period of the adolescent to adult transition, relationships also vary in form or type, as some individuals move in together or get married, and others continue a pattern of dating. Some researchers have also documented an association between the form of relationships and IPV with cohabiting unions demonstrating higher odds of IPV [11]. The current investigation takes into account basic differences in partner continuity, as well as relationship type and their association with IPV as observed across the study period. Most important, however, is the need to capture variations in relationship qualities, as specific dynamics within the relationship are likely to represent concrete, proximal sources of variations that are directly implicated in IPV experience.

Previous cross-sectional assessments have documented that infidelity, relationship churning (breaking up and reconciling), and verbal conflict/disagreement are all associated with higher odds of reporting IPV [12, 13]. Lower levels of positive relationship dimensions such as trust or partner validation have also been associated with higher risk [12]. However, it is important to determine whether this reflects a kinds-of-people association, or intrapersonal change. A longitudinal lens is important because cross-sectional portrayals can lead to unwarranted conclusions such as once an abuser or victim, always an abuser or victim [14]. This contrasts with basic tenets of the life course perspective, which accommodates the idea that experience within romantic relationships themselves can be the basis for further learning and adjustment [15]. Previous longitudinal studies have enhanced understanding of between-individual differences [1, 13, 17, 18], but continue to shed little light on within-person change. A more nuanced analysis of how individuals navigate relationships from adolescence to adulthood is warranted to capture discontinuities in IPV experiences.

Although many intimate partner violence experiences reported in population-based samples are mild (pushing, shoving, grabbing), they often involve both partners’ use of violence, and can produce injuries [19]. Yet, many studies limit analysis (due to data constraints) to only victimization or only perpetration. Acknowledging that bidirectional violence often occurs [15, 20] and that patterns of victimization and perpetration may shift over time, the current study examines patterns of any IPV (victimization and/or perpetration), recognizing the need to build on this initial portrait with more nuanced investigations of various types and levels of IPV experience.

Drawing on longitudinal population-based data, the Toledo Adolescent Relationships Study (TARS), we provide a descriptive portrait of individuals’ proportion of relationships with IPV from adolescence into adulthood, and fixed effects analyses examining how changes in relationship context relates to the accumulation of IPV over time. We hypothesize that such intrapersonal changes will reflect a complex risk profile that may increase or decrease IPV risk. Specifically, increases in infidelity, relationship churning, frequency of disagreements, and partner continuity will be associated with increases in the proportion of relationships with IPV, while trust, validation, self-disclosure, and commitment will be associated with decreases in the proportion of relationships with IPV.

Methods

Data

The TARS sample (n = 1,321) was drawn from the year 2000 enrollment records of all seventh, ninth, and eleventh graders in Lucas County, Ohio. The sampling frame, devised by the National Opinion Research Center, consists of a stratified, random sample of students enrolled in school (but not necessarily attending) based on grade, race-ethnicity, and gender. The TARS sample of 1,321 is 81.3 percent of the original 1,625 students who were contacted. We conducted interviews in respondents’ homes using preloaded laptops to maintain privacy and respondents received gift cards in the amount of $25 (interviews 1 and 2), $50 (interviews 3 and 4), and $75 (interviews 5) as compensation. This study was reviewed and approved by Bowling Green State University’s institutional review board.

We drew on data across all five interviews or “waves.” Wave 1 was conducted in 2001, wave 2 in 2002–2003, wave 3 in 2004–2005, wave 4 in 2006–2007, and wave 5 in 2011–2012. Retention rates from the first to second interview were 89.1%, 84.4% for the third interview, 82.8% for the fourth interview, and 77.8% for the fifth interview. Participation rates were lower for Black youth, males, and those who changed partners between waves. Reports of any IPV were not contingent on participation rate. To assess variation in the proportion of IPV reports across relationships, we restricted the fixed-effects analysis to those with two or more partners, (n=1,146 subjects, 3,534 observations). The vast majority, 87%, reported two or more relationships from adolescence to young adulthood. An advantage of the analytic strategy (described below) is that it is not necessary to restrict the analytic sample to those with complete data at all five waves. Our multilevel modeling approach is able to use any available data at level one (within-person), but must have complete data at level two (between-person). With less than 0.025% (n = 3) missing data at level two, we chose to delete these observations from the analytic sample.

Measures

Variable distributions are presented in Table 1. Prior longitudinal assessments of IPV have focused on continuity across two time points, and whether IPV continuity differed for those who changed versus stayed with the same partners [1, 20]. We assessed IPV frequency across changing and continuing relationships using a measure that accounted for the number of relationships with IPV relative to the total number of relationships accumulated at each of the five interviews. This is important because a respondent who reported IPV twice across five relationships would suggest greater discontinuity than a respondent who reported IPV twice for only three relationships. Thus, the IPV measure assessed the proportion of relationships with any IPV perpetration or victimization. We combined perpetration and victimization experiences because preliminary analyses revealed that the majority of those reporting IPV at one time period were likely to change in their IPV status or form at the next period. For example, examination of patterns between waves 4 and 5 revealed that three-quarter of respondents (74.7%) who reported IPV at wave 4 changed IPV status (yes/no) or form (perpetration/victimization) at interview 5. Furthermore, although the issue of gender symmetry in intimate partner violence remains controversial, studies based on community samples have reported high rates of bidirectional violence among youth [18, 21]. Therefore, this measure reflects that bidirectional violence often occurs and that the form and status of IPV is likely to change across adolescence and young adulthood.

Table 1.

Descriptive statistics for covariates by age group (N = 1,146 subjects, 3,534 observations).

Variable	Age group 1 13 – 16 years (n = 593)		Age group 2 17 – 20 years (n = 1528)		Age group 3 21 – 24 years (n = 847)		Age group 4 25 – 29 years (n = 566)
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Infidelity	0.43		0.43		0.38		0.35a
Relationship churning	0.53		0.52		0.50		0.45ab
Frequency of disagreements (1 – 5)	2.26	(0.97)	2.62a	(0.99)	2.72a	(0.96)	2.73a	(0.90)
Trust (1 – 5)	3.74	(1.15)	3.72	(1.21)	3.86b	(1.19)	4.06abc	(1.18)
Validatio (1 – 5)	3.88	(0.82)	3.99a	(0.79)	4.03a	(0.79)	3.99	(0.85)
Self-disclosure (1 – 5)	3.64	(1.01)	3.82a	(1.00)	3.94ab	(0.91)	4.04ab	(0.84)
Commitment (1 – 5)	3.92	(1.07)	3.91	(1.03)	3.78b	(1.10)	3.94c	(1.11)
Partner c	0.11		0.21ac		0.29ab		0.30abc
Relationship Type
Dating	1.00		0.88ac		0.63ab		0.38abc
Cohabiting	0.00		0.10ac		0.27ab		0.35abc
Married	0.00		0.01		0.10ab		0.27abc
Parenthood	0.02		0.12ac		0.33ab		0.45abc
Gainful activity	0.98		0.70a		0.67a		0.69a
Receipt of public assistance	0.09		0.17ac		0.29ab		0.39abc
Antisocial behavior (0 – 1.75)	0.17	(0.31)	0.24a	(0.34)	0.27ab	(0.33)	0.19bc	(0.27)
Age	15.38	(0.83)	18.64	(1.03)	22.36	(1.03)	26.48	(1.15)

^asignificantly different from “age group 1”;

^bsignificantly different from “age group 2”;

^csignificantly different from “age group 3”;

Note: All significance levels at p < .05.

Source: Toledo Adolescent Relationships Study

Any IPV was assessed for each current/most recent relationship using four items from the Conflict Tactics Scale [22]. Respondents were asked how often they committed the following acts against their current or most recent partner or their partner committed these acts against them: “thrown something”; “pushed, shoved or grabbed”; “slapped in the face or head with an open hand”; and “hit.” Responses ranged from 1 (never) to 5 (very often). Respondents who answered affirmatively to any item were coded as 1 and 0 otherwise. Alphas ranged from .89 to .91. Total number of IPV reports was based on the number of relationships in which respondents reported IPV. Proportion of IPV reports was the number of relationships with IPV accumulated at each wave, divided by the number of partners accumulated by that same wave. This created a dynamic proportional measure that varied over time based on the number of IPV reports and number of partners. Thus, rather than assuming that IPV exposure demonstrates a positive linear trend over the course of adolescence and young adulthood (as it would be with a raw count of IPV experiences), values may increase or decrease as a function of the number of IPV reports relative to the number of relationships overall. This measure accounts for the possibility of movement into new relationships not characterized by violence, which also influences the overall cumulative risk. Proportional values ranged from 0.0 (no relationships with IPV) to 1.0 (IPV reported for all five relationships).

The fixed effects models included time-varying variables in assessing whether changes in the romantic context are related to the proportion of IPV reports. Infidelity was based on two items asking how often respondents or partners “saw another girl/guy,” and “were physically involved with other girls/guys.” Alphas ranged from .72 to .83. Relationship churning was coded 1 if respondents reported breaking up and reuniting with a partner at least once, and 0 otherwise. Frequency of disagreements was measured by asking how often respondents and partners disagreed or argued. Responses ranged from 1 (never) to 5 (very often). Trust was assessed by asking the extent to which respondents could trust their partners. Responses ranged from 1 to 5 with higher scores reflecting greater trust. Validation was the mean of two items: “[partner] makes me feel attractive”; and “[partner] makes me feel good about myself.” Responses ranged from 1 (strongly disagree) to 5 (strongly disagree). Self-disclosure assessed how often respondents talked to partners about the following: “something really bad happened,” “home and family life,” and “private thoughts and feelings.” Responses ranged from 1 (never) to 5 (very often), and alphas ranged from .83 to .89 for the mean scale. Commitment was assessed with the item: “how often have you seriously considered ending your relationship with [partner]?” Responses ranged from 1 (never) to 5 (very often). Partner continuity was coded 1 if respondents were with same partners between waves, and 0 if respondents reported new partners. Relationship type assessed whether respondents were in dating (referent), cohabiting, or marital unions.

Developmental factors associated with transitioning to adulthood and related to IPV were operationalized as time-varying. Parenthood was coded 1 if respondents reported biological children, or children in the household. Gainful activity was coded 1 if respondents were attending school or employed full-time, and 0 otherwise. Receipt of public assistance was coded 1 if respondents reported receiving government or public assistance, and 0 otherwise. Antisocial behavior was a logged mean scale of 9 items from a 26-item inventory [23]. Items asked how frequently respondents engaged in theft (major and minor), breaking and entering, assault and battery, property damage, selling drugs, carrying a hidden weapon, public drunkenness, and drug use. Alphas ranged from .74 to .88.

Analysis

To examine associations between changes in the romantic context and proportion of IPV relationships, we used Allison’s hybrid method, which separates each of the time-varying predictors into two components – a within-person variation component and a between-person variation component [24]. By group-mean centering the time-varying predictors this modeling strategy allows us to produce fixed effects in a random effects model. Thus we focus on within-person variation while controlling for unobserved heterogeneity (i.e., selection factors) that potentially may be related to both IPV and developmental changes [25, 26]. This method offered the advantage of assessing whether any of the effects varied by gender since we cannot introduce a time-stable predictor in a traditional fixed-effects model. A Hausman test of differences in the coefficients between random-effects and fixed-effects models supported our focus on the fixed-effects coefficients. Given our developmental focus and that previous research has demonstrated that IPV displays an age-specific pattern [27], we use age as our measure of time. Due to the unbalanced nature of our data, respondents limited to a single observation contributed only to the intercept, because no within-person variation exists [28]. This method produced coefficients for within-person change and between-person coefficients. Since our focus is on within-person change we report the individual level change coefficients.

Results

Table 2 provides a distribution of the proportion of relationships with any IPV by the time the sample had reached wave 5 for men and women, collapsing the proportional measure into three categories – no IPV, IPV in some relationships, and IPV in all. For the total sample, continuity is represented by the 38.9% who had no IPV and 7.5% who reported always having IPV. Half (53%) experienced discontinuity in IPV and among those reporting violence, the overwhelming majority (87%) reported some level of discontinuity. T-tests indicated similar shares of men and women experienced discontinuity in violence.

Table 2.

Proportion of Relationships with Any Intimate Partner Violence among Young Adult Men and Women at Wave 5, Toledo Adolescent Relationships Study.

		Full n = 1146			Men n = 557 / 48.6%			Women n = 589 / 51.4%
		Total	w/IPV		Total	w/IPV		Total	w/IPV
No IPV	446	38.9%		181	32.5%		265	45.0%
IPV in some relationships	609	53.1%	87.0%	329	59.1%	87.5%	280	47.5%	86.4%
IPV in all relationships	91	7.9%	13.0%	47	8.4%	12.5%	44	7.5%	13.6%

We first ran an unconditional means model to calculate the intraclass correlation coefficient (ICC) in order to assess how much of the total variation of IPV is within individuals versus between individuals. With an intercept of 0.06365 (z = 21.48, p < .0001) and a residual of 0.02171 (z = 34.76, p < .0001), 75% (0.06365 / 0.06365 + 0.02171) of the variation in IPV is at the within-person level further supporting our focus on change.

Figure 1 presents unconditional growth models by gender. The linear and quadratic effects of age were significant for each of the models. Thus, the proportion of relationships with IPV increased through adolescence, and peaked during the early twenties, followed by a decline in the proportion in the mid- and late-twenties. The age terms did not interact with gender, indicating that men and women’s patterns were similar.

Figure 1.

Predicted proportion of relationships with IPV by age and gender.

Table 3 presents the results of the fixed-effects analysis. Consistent with our hypotheses, increases in infidelity and frequency of disagreements relative to the individual’s average scores were associated with greater proportions of relationships with IPV. Increases relative to individual averages in trust and commitment were associated with a lower proportion of relationships with IPV. These results suggested that relationships characterized by greater trust, commitment, fidelity, and less verbal conflict corresponded to decreases in IPV experiences. Retaining a partner between waves was associated with a higher proportion of relationships with IPV, indicating that partner turnover is associated with lower accumulation of IPV. Cohabitation relative to dating was associated with an increase in the proportion of relationships with IPV. Finally, both of the age terms remained significant.

Table 3.

Fixed Effects Model for Proportion of Relationships with Any Intimate Partner Violence across Adolescence and Young Adulthood (N = 1,146 subjects, 3534 observations).

	Coef.	S.E.
Intercept	.173*	.087
Infidelity	.020**	.007
Relationship churning	.012	.007
Frequency of disagreements	.020***	.004
Trust	−.008*	.003
Validation	−.003	.004
Self-disclosure	.003	.004
Commitment	−.010**	.004
Partner continuity	.047***	.008
Relationship Type (Dating)
Cohabiting	.019*	.004
Married	.024	.014
Parenthood	.004	.011
Gainful activity	.005	.007
Receipt of public assistance	.016	.009
Antisocial behavior	.013	.013
Age (centered at age 13)	.010**	.003
2 Age(centered at age 13)	−.001**	< .001

^*p < .05;

^**p < .01;

^***p < .001

Note: Between-subject effects are included in the model but not shown

We tested whether any of the covariates interacted with age or gender. None of the interactions with age at the within-person level were significant. Self-disclosure and infidelity both interacted with gender at the within-person level such that changes in each were associated with stronger effects for women.

Discussion

This study revealed greater variability than stability in IPV across adolescence and young adulthood for men and women. More than half of respondents reported at least one IPV experience, but for most this was not representative of their relationships overall. Less than one in 12 (8%) reported experiencing IPV in all of their relationships. Thus, consistent with prior work, we show that IPV is quite prevalent among youth. Yet, results also highlight that IPV experiences demonstrate a high degree of variability.

Consistent with recent work demonstrating that most young adults move toward IPV desistance [15, 29], age demonstrated a curvilinear pattern with increases in the proportion of relationships with IPV during late adolescence followed by decreases in young adulthood. Results suggested that relationships not only increased in trust and intimacy, but also churning occurred less often as youth moved from dating into more committed unions of cohabitation and marriage. It appears, however, that these transitions may be accompanied by increases in the frequency of disagreements. Disagreements are not necessarily detrimental [30], but the nature of conflicts and how partners manage conflicts likely have a stronger influence on IPV. Nevertheless, increases in disagreements corresponded to a proportional increase in the number of relationships with IPV. Increases in trust and commitment were associated with a lower proportion of relationships with IPV. Thus, higher quality romantic relationships were associated with a lower accumulation of IPV experiences. These findings support the notion of a relationship learning curve [15] in which young adults draw on the full breadth of relationship experiences to inform choices and conduct in the romantic domain. Such choices may involve developing criteria for the selection of a new partner, and creating boundaries of what will or will not be tolerated within a relationship.

The relationship factors were associated with IPV in a similar manner for men and women, except for self-disclosure and infidelity. The within-person interactions showed that associations between self-disclosure and IPV and infidelity and IPV were greater for women than men. This suggests that changes in self-disclosure or infidelity may reflect relationship problems that are specific to women that result in increased risk for relationship aggression.

Consistent with our expectations and previous work [31], partner continuity was associated with increases in the proportion of relationships with IPV. Paradoxically settling down with a single partner would appear to reflect healthy development and a stronger bond, and yet retaining the same partner provides more opportunity for disagreements and conflict to occur. In general, as youth move into their twenties, this analysis and previous research document declines in the prevalence of IPV among young adults. It is possible, however, that the association between partner continuity and IPV reflects that some youth respond to the presence or threat of IPV by leaving the relationship. We considered the possibility that partner continuity may reflect a risk during adolescence (settling down too soon), but is a protective factor as youth get older. Yet, the interaction between age and partner continuity, was not significant. Efforts are needed to further understand the contrasting risk and rewards associated with settling down. Future work should include consideration of partner characteristics and information elicited from both members of the couple.

The current study has some limitations. Ideally, changes in IPV over the course of a relationship would have been assessed, but questions asked how often respondents experienced IPV in their relationships rather than changes within relationships. Although some respondents retained partners between waves and reported IPV at each wave, we cannot be sure that this represents different IPV occasions, or a single IPV event. We also recognize that our combined measure of IPV may mask important differences between victimization and perpetration experiences. Given that some populations are particularly at risk for IPV [32, 33, 34], future work should strive to untangle these intricacies. Additionally, our measure did not assess severity, frequency, or injury. Since women experience greater fear and more injuries as a result of IPV [35], this is an important consideration. Finally, although TARS includes some individuals in same-sex relationships, their numbers were too small for separate analyses.

Conclusions

The current study demonstrates that IPV, although prevalent among youth, does not represent a consistent experience. Policymakers, however, should exercise caution in efforts to target specific individuals for prevention and intervention. The possibility of stigmatizing youth as perpetrators or victims has the potential to ultimately inhibit help-seeking behavior [36] and possibly produce other unforeseen harm. Our findings support previous calls for greater emphasis on education promoting the development of healthy relationships among adolescents and young adults, as the findings underscore that relationship-specific factors make a difference, even after sociodemographic and adult status characteristics as well as lifestyle factors (antisocial behavior) have been taken into account.

Implications and Contribution.

Intimate partner violence (IPV) experiences varied substantially across adolescence and young adulthood, with the majority reporting discontinuity in IPV across relationships. Improvements in relationship quality were associated with a lower accumulation of IPV experiences further reinforcing calls for programs that emphasize building healthy relationships during adolescence and young adulthood.