Financial Behaviors, Couple-Level Conflict, and Adolescent Relationship Abuse: Longitudinal Results from a Nationally Representative Sample

Abstract

Financial disagreements have been identified as a severe source of discord in adult relationships, yet limited work has considered whether financial considerations contribute to conflict among younger samples. Drawing on longitudinal data from a nationally representative sample of adolescents, the current investigation examined the extent to which money lending practices, feelings of financial resentment, and exposure to economic control contribute to couple-level interactions, and in turn, to conflict that escalates to the point of violence. Findings provide evidence of an association between adolescent financial behaviors and concurrent conflict due to economic considerations. Moreover, conflict due to economic considerations was an important predictor of future adolescent relationship abuse perpetration. We discuss the implications of our findings for intervention/prevention efforts.

Intimate partner violence (IPV) has been identified as a significant public health problem with deleterious consequences for individuals’ immediate and long-term well-being. Although the majority of researchers have focused on violence in adult relationships, there is a burgeoning body of literature on violence in adolescent relationships. Findings from this work have revealed that adolescent relationship abuse (ARA) is a serious problem. According to recent estimates, physical violence occurs in up to two-fifths of dating relationships (Mumford & Taylor, 2016; Sears, Byers, & Price, 2007; Silverman, Raj, Mucci, & Hathaway, 2001). Indeed, examination of longitudinal patterns reveals that rates of partner violence increase across the adolescent period, reaching a peak during young adulthood (i.e., between the ages of 17–20 for male youth and 21–24 for female youth) (Johnson et al., 2015). Furthermore, the experience of relationship violence in adolescence has consequences for psychosocial development, as well as mental and physical health outcomes (e.g., Ackard, Eisenberg, & Neumark-Sztainer, 2007; Copp, Giordano, Longmore, & Manning, 2016a; Exner-Cortens, Eckenrode, & Rothman, 2013; Haynie et al., 2013). Preliminary research on youth dating violence has helped to identify risk factors (e.g., Foshee et al., 2011; Giordano, Soto, Manning, & Longmore, 2010; O’Keefe, 2005; Vagi et al., 2013), and has guided the development of intervention and prevention programming (Coker, Banyard, & Recktenwald, 2017; Cornelius & Resseguie, 2007; Taylor, Stein, Mumford, & Woods, 2013). However, there is a lack of research on the mechanisms driving the use of violence in these early dating relationships.

In the adult literature, a number of theoretical frameworks have been used to examine the etiology of IPV, including social learning approaches, feminist perspectives, and theories of individual difference (e.g., attachment theory). The adolescent dating violence literature, in contrast, is largely atheoretical. Thus, the prevailing strategy is to draw on frameworks developed to understand adult intimate partner violence (Mulford & Giordano, 2008). There are, however, a couple of important exceptions. For example, Riggs and O’Leary (1989) developed a model of courtship aggression, which posited that the predictors of dating violence can be separated into background factors (i.e., factors establishing an individual’s proclivity toward violence) and situational factors, or “triggers” (i.e., drinking, relationship problems). Others have adopted a more developmental perspective (e.g., Exner-Cortens, 2014; Wekerle & Wolfe, 1999), incorporating theories of adolescent development and prior research on dating violence. From this work, we can glean the following about ARA: 1) a comprehensive model of adolescent partner violence requires attention to multiple levels of risk (i.e., individual and relational/situational factors) and 2) research on partner violence during adolescence must account for unique developmental features.

Despite a recent push to develop unique theories of adolescent relationship violence, we suggest that there may be some utility in attending to relationship dynamics and other risk factors that may be related to partner violence among both adolescents and adults, as well as unique features of the adolescent period. In the current investigation, we draw on data from the National Survey on Teen Relationships and Intimate Violence (STRiV) to examine associations between economic considerations and ARA within the dyadic context. Our focus on economic considerations, including fighting about money issues, the provision of economic support, feelings of financial resentment, and exposure to economic control, is motivated by prior research on partner violence among adults. Numerous studies have revealed that issues of financial support and economic dependence are risk factors for IPV, as well as barriers to leaving relationships that include violence (Hendy, Eggen, Gustitus, McLeod, & Ng, 2003; Strube & Barbour, 1983). Thus, leveraging the longitudinal design of the STRiV data, the goal of the current research is to determine whether this particular domain (i.e., economic considerations) is predictive of subsequent violence in adolescent relationships. In addition, we attempt to unpack the relationship dynamics associated with economic conflicts within this demographic group.

Economic Considerations, Relationship Conflict, and IPV/ARA

In the IPV literature, several studies have examined the link between objective markers of economic disadvantage and partner violence (e.g., Cascardi & Vivian, 1995; Cunradi, Caetano, Clark, & Schafer, 2000; Jasinski, Asdigian, & Kantor, 1997), suggesting that the stresses associated with poverty contribute to relationship conflict. This is consistent with the findings of family scholars, who have noted the impact of economic problems on family stress and hostile patterns of interaction between partners (Edin & Kissane, 2010; Kinnunen & Pulkkinen, 1998). In fact, financial disagreements are often more severe than other areas of discord, thus elevating the risk of violence (Dew & Dakin, 2011; Papp, Cummings, & Goeke-Morey, 2009). Although the link between economic stressors and relationship processes (including violence) is well-established, prior work does not provide a window to the mechanisms through which financial considerations influence interactions at the couple-level. Furthermore, the majority of prior research has focused on adults. This is somewhat limiting as the norms regarding economic matters vary greatly across relationship types and developmental periods.

Notably, there is some preliminary evidence to suggest that economic considerations have consequences for relationship dynamics at earlier points in the life course. For example, Copp, Mumford, and Taylor (2016) found that requests made by a partner to borrow money (i.e., the practice of money lending in dating relationships) was associated with ARA perpetration and victimization among a nationally representative sample of adolescents. In addition, recent research has linked financial conflicts to IPV risk among a large, diverse sample of young adults (Copp, Giordano, Manning, & Longmore, 2016)--underscoring the potential utility of looking beyond general discord to a more nuanced understanding of the specific content of couples’ disputes. An important limitation of this prior work, however, is that it was cross-sectional. Accordingly, it is important to further examine these patterns longitudinally to determine whether economic considerations influence the use of violence within the dyadic context.

That prior work has consistently identified a link between financial difficulties and couple conflict suggests a need to consider whether certain relationship skills/deficits influence the extent to which adolescent couples are able to successfully negotiate conflict stemming from difficult economic circumstances. In the current investigation, we direct attention to a range of relationship dynamics, including conflict about financial issues, money lending, financial resentment, and economic control/influence. While prior research indicates that the majority of youth learn financial knowledge from parents through a process of financial socialization (Chen & Volpe, 2002), adolescent romantic relationships provide an early context for individuals to begin to manage basic financial matters. Although adolescents’ financial socialization/literacy and other economic considerations may not be particularly strong predictors of ARA given the relatively limited financial responsibilities of youth during this phase of the life course, these are important skills for adolescents to develop as financial decision-making is a particularly important domain of contestation if and when couples face real economic hardship and other potentially emotionally laden family choices (Papp et al., 2009). Thus, in addition to considering associations between economic considerations and ARA perpetration, we consider whether such factors contribute to concurrent conflict about money within the relationship.

Prior Theorizing and Research on Partner Violence

Research on the etiology of IPV has often emphasized social learning mechanisms. According to a social learning approach, behaviors are learned through the observation and imitation of others (Bandura, 1978). Early interactions with parents are particularly formative, from a learning perspective, and teach children a variety of behaviors. Many investigations have examined the relationship between violence in the family of origin, including witnessing interparental violence and experiencing child abuse, and partner violence (e.g., Ehrensaft et al., 2003; Foshee, Linder, MacDougall, & Bandiwala, 2001; Kitzmann, Gaylord, Hold, & Kenny, 2003; Kwong, Bartholomew, Henderson, & Trinke, 2003; White & Widom, 2003). Although social learning theory remains one of the most popular explanations of relationship violence, research on the intergenerational transmission of partner violence indicates that associations between early exposure to violence and subsequent involvement in violent relationships are modest (Stith et al., 2000). Furthermore, research adopting a social learning approach to partner violence is often somewhat unsophisticated, as scholars tend to focus on the direct transmission of certain behaviors (i.e., imitation). Yet there is more to social learning than imitation; individuals observe and may internalize complex scripts for violent behavior, verbal tactics and styles of communication, and attitudes accepting of violence (Copp, Giordano, Longmore, & Manning, 2016b). Consistent with prior research, we include in our models indicators of violence in the family of origin. In addition, consistent with social learning theory, we consider the role of attitudes toward partner violence and communication processes as correlates of economic conflicts, as well as predictors of ARA perpetration in their own right.

Feminist theories are also commonly applied to the study of partner violence. These perspectives direct attention to the role of gender within adult relationships, with a particular focus on dynamics of power and control (Dobash & Dobash, 1979). A feminist analysis posits that violence, and related control attempts, represents an assertion of male dominance. Because feminist approaches view relationship violence as an outgrowth of patriarchal values and gendered power systems, control is most often examined with respect to male-to-female partner violence. However, control has also been linked to female violence perpetration (Giordano, Copp, Longmore, & Manning, 2015; Stets & Pirog-Good, 1989), suggesting that the traditional feminist view may be incomplete as an explanation. In a recent article, Giordano and colleagues (2015) indicated “the overarching goal of control or dominance may not fully capture the range of relationship dynamics associated with IPV” (p. 924). Building on this work, we move beyond the notion that violence is the outgrowth of a more general male desire to control female partners, and focus instead on specific relationship dynamics and patterns of interaction. We argue that this focus may be of particular import during the adolescent period, where female perpetration rates are as much as three times that of their male counterparts (Taylor & Mumford, 2016). Nevertheless, we recognize the important contribution of feminist perspectives to our understanding of potential gender differences in the pathways and consequences of partner violence, and pay specific attention to gender throughout the analyses. In particular, evidence of a gender gap in financial literacy (Chen & Volpe, 2002; Fonseca, Mullen, Zamarro, & Zissimopoulos, 2012) underscores the multiple layers through which mechanisms leading to violent relationships may differ by gender.

Another line of theorizing adopts an individualized psychological approach, focusing on personality traits, attachment styles, or general propensities toward violence. For example, Moffitt and colleagues (2000) found evidence of a link between partner abuse and negative emotionality. A typology of male batterers developed by Holtzworth-Munroe and Stuart (1994) uses indicators of antisociality and borderline personality features to classify maritally violent men. Felson and Lane (2010) determined that men and women who engage in intimate partner violence are similar to other types of violent offenders, as both have a predisposition toward violence. These are just a few of the more individualized approaches to IPV (see also e.g., Ehrensaft, Cohen, & Johnson, 2006; Wekerle & Wolfe, 1999; White & Widom, 2003). A shortcoming of approaches that focus on individual differences is that they are ill-equipped to explain variation in partner violence, including cessation within a particular relationship, discontinuity in the use of violence across relationships, and variability at different stages of development (Capaldi & Kim, 2007; Capaldi, Shortt, & Crosby, 2003; Giordano et al., 2015). Nevertheless, individual differences, including violent tendencies, are robust predictors of IPV. Accordingly, we control for general aggression in multivariate analyses.

Informed by these rich theoretical traditions, the current investigation applies a social learning lens that focuses on the formative influence of family background factors, while also directing attention to process-level variables to understand the situated nature of violent couple-level interactions. This is consistent with other recent conceptual models that focus on dyadic factors and interactional processes (e.g., Capaldi & Kim, 2007; Giordano et al., 2015; O’Leary & Slep, 2010). In addition, we draw on elements of feminist and individualized approaches to further develop our understanding of predictors of ARA, as well as the correlates of specific areas of disagreement (i.e., fighting about money) that may escalate to violence in adolescent relationships. In particular, we account for family background, individual, and relational/situational factors in two sets of analyses, including models examining: 1) the correlates of conflict due to economic considerations and 2) estimating the odds of ARA perpetration. This approach is further informed by attention to unique features of the adolescent period.

Unique Features of the Adolescent Period

Adolescence is characterized by a marked shift in interpersonal relationships, as reliance on parents begins to wane and romantic partners take on increasing importance (Collins, 2003). Participation in romantic relationships increases across this developmental period, and by 18 more than 70% of individuals report some relationship involvement (Carver, Joyner, & Udry, 2003). Although adolescent relationships are often viewed as superficial, insignificant, and short-lived, research has underscored the developmental significance of these early partnerships (Collins, 2003). In fact, prior work indicates that by the end of adolescence, partners become adolescents’ primary source of love, support, and interdependence (Laursen & Williams, 1997). Further, over the course of adolescent development, romantic relationships increasingly include behaviors characteristic of adult relationships such as the provision of support, affection, and nurturance (Furman & Wehner, 1994), while preserving other behaviors more typical of adolescent peer relationships.

Notably, the skills acquired in adolescent relationships, including the provision of support, communication, and conflict resolution techniques, seem to carry over to future relationships (Collins, Welsh, & Furman, 2009). In this way, healthy adolescent romantic relationships provide an important context for the development of future healthy relationship behaviors. In a similar vein, adolescent relationships may encourage the development of unhealthy relationship behaviors, including the use of physical violence, which heightens youths’ risk of involvement in future relationships that include violence (Gomez, 2010; Taylor & Mumford, 2016). Taken together, these findings suggest that the relational patterns established in adolescence may carry over into adulthood, thus underscoring the need to intervene before problem behaviors/dynamics become entrenched. And in fact, a number of programs have been developed to help encourage healthy relationships among adolescents (see Niolon et al., 2016), with a focus on relationship skill-building in areas including respect, consent, and communication (see De Koker, Mathews, Zuch, Bastien, & Mason-Jones, 2014 for a review). Although dating violence and its warning signs are often discussed, there is limited (albeit increasing) attention to the dyadic nature of IPV and the specific content of partner disputes. Relationship ‘warning signs’ or stressors may be interpreted by adolescents in relatively abstract terms, and thus it is potentially useful to anchor these concepts to adolescents’ specific life course circumstances that contribute to relationship conflict.

For example, an explicit focus on economic control as a risk factor for ARA within intervention/prevention programming may not resonate with youth, as these early relationships do not include the same level of economic dependency or shared finances as adult relationships. In fact, particularly in early adolescence, romantic partners are unlikely to share any financial responsibilities. However, adolescent dating partners do begin to provide each other with instrumental support, by paying for meals or shared activities, purchasing gifts, and helping out with expenses (Giordano et al., 2010; Kuttler & La Greca, 2004; O’Sullivan & Meyer-Bahlburg, 2003). As a result, there is the potential for one member of the couple to perceive that he/she is contributing more financially to the relationship, which may foster feelings of resentment. Furthermore, as adolescent relationships progress and romantic partners become increasingly enmeshed, couples may begin to call on one another for direct financial assistance and/or become more involved in each other’s financial decision-making. Given that economic factors are such a common source of relationship conflict, it is important that youth learn to navigate financial considerations early on in order to establish boundaries within their relationships, and to gain experience negotiating these types of relationship matters.

The Current Study

In the current investigation we focus on a particular domain, economic factors, in order to unpack the extent to which such considerations contribute to couple-level interactions, and in turn, to conflict that escalates to the point of violence, among a nationally representative sample of adolescents. Linkages between economic considerations, including economic control, and IPV are well-documented in prior research using adult samples (e.g., Cunradi et al., 2000; Golden, Perreira, & Durrance, 2013; see also Capaldi, Knoble, Shortt, & Kim, 2012 for a review). In addition, a limited number of studies have examined associations between economic factors and partner violence during the adolescent and young adulthood periods of the life course (see e.g., Copp, Giordano, Manning, & Longmore, 2016; Copp, Mumford, & Taylor, 2016; Giordano et al., 2015). To date, however, research examining the role of economic factors in ARA has been cross-sectional.

In the current investigation, we contribute beyond prior work by examining these associations longitudinally, and exploring risk factors for negative forms of communication with respect to economic considerations in particular. A goal of the current investigation was to describe the types of economic considerations present in adolescent relationships and, furthermore, to illustrate associations and longitudinal linkages between financial support, feelings of resentment, economic control, conflict due to money, and the perpetration of ARA. Few studies have examined the role of economic control using younger samples. In an exception, Copp, Mumford, and Taylor (2016) examined associations between financial behaviors and adolescent dating violence, and found that requests for instrumental support from a romantic partner (i.e., money lending) were associated with heightened risk of moderate and serious threats/physical violence. The current study extends this work by incorporating a broader range of financial considerations and examining associations between these factors and ARA longitudinally. Consistent with prior research, we control for a range of traditional predictors and sociodemographic characteristics. Based on the prior research and theorizing, we propose the following hypotheses:

H1: Adolescent financial factors (i.e., financial resentment, money lending, economic control, and socialization) will be significantly associated with concurrent relationship conflict due to economic considerations net of traditional predictors and a range of sociodemographic characteristics, such that financial resentment, money lending, and economic control are associated with heightened risk, and socialization with lower risk, of conflict.

H2: Adolescent financial factors, including economic conflicts, will be significantly associated with subsequent physical violence perpetration. In particular, economic conflict, financial resentment, money lending, and economic control will be positively associated with physical violence perpetration, and socialization will be negatively associated with physical violence perpetration.

H3: The observed associations between adolescent financial factors and physical violence perpetration will differ for men and women. Specifically, given differences in traditional gender role expectations with respect to financial provision/decision-making, we expect the associations between financial behaviors and violence to be stronger for men.

Data and Methods

Data for this study were drawn from the STRiV study, waves 1 through 3, collected annually. STRiV respondents were recruited from the GfK KnowledgePanel, a national household probability sample of the United States (a full description of the panel is available at www.knowlegenetworks.com/knpanel/doncs/KnowledgePanel(R)-Design-Summary-Description.pdf). From the KnowledgePanel, a sample of 2,354 pairs of one parent/caregiver and one random co-resident child between the ages of 10–18 years old was recruited (collected over the period October 2013 to January 2014). Of the baseline adolescent respondents, 1,471 (62.5 %) completed the wave 2 survey and 1,553 (66.0%) completed the wave 3 survey. We used the same online survey methodology for the collection of confidential and protected data for the parent and youth data collection across all three survey waves (see Taylor & Mumford, 2016, for more detail on the methods used in this study).

When recruited to the KnowledgePanel, sampled households not connected to the Internet are provided a netbook computer and free Internet service. All contacts (via telephone, postcard, etc.) and survey materials are available in both English and Spanish. There were no substantive demographic differences between the STRiV sample and national estimates of race, biological sex, socioeconomic status (SES), region, and urbanicity (Taylor & Mumford, 2016). Following standard survey and IRB protocols, recruited respondents were informed in advance and within the online survey that they could refuse to participate or, if participating, refuse to answer any questions. Analyses revealed no apparent patterns to survey non-participation. Panel demographic post-stratification weights were applied to adjust both for non-coverage of the U.S. population as well as participant non-response and missing data. Further detail on the Knowledge Panel and the STRiV sampling and recruiting methods is available (Taylor & Mumford, 2016).

The sample for the current investigation includes adolescent respondents who participated in all three interview waves and reported at both waves 2 and 3 that they were currently in a dating relationship or had been in a dating relationship that had lasted at least a week within the past year, and had valid responses on the outcome variable, ARA perpetration (n = 379 after applying this criteria). Attrition analyses revealed that the analytic sample was similar to the baseline sample across most study variables (those retained were more likely to report witnessing parental violence and general aggression, were slightly older, more likely to be non-Hispanic white, and less likely to report having college-educated parents). In addition, Heckman two-step correction models were employed to consider the potential for selection bias. Findings suggested that selection bias does not exist in our data (inverse Mills ratio (IMR) = 0.016, p = 0.984). These data contain very few missing cases, with less than 2% missing for most variables included in the current investigation (one variable was missing 11% of observations). Despite the small number of missing cases, missing data was handled using multiple imputation.

Measures

Adolescent Relationship Abuse Perpetration (collected during the wave 3 survey)

We used a modified version of the Conflict in Adolescent Dating Relationships Inventory (CADRI) to assess adolescent relationship abuse perpetration in the sample (Wolfe, Scott, Reitzel-Jaffe, Wekerle, Grasley, and Straatman, 2001). During the wave 3 interview, respondents were asked whether they had ever engaged in a range of violent behaviors within the context of a current or most recent dating relationship (e.g., kicking, hitting, punching, scratching, bending fingers, slapping, pulling hair, pushing, shoving, shaking, burning, choking, biting, and using a knife or gun against victim). Based on these responses, we created a dichotomous measure of adolescent relationship abuse perpetration, distinguishing between those who reported any (1) and no perpetration (0).

Adolescent Financial Factors (wave 2)

Several financial measures were developed and piloted for the STRiV study. We drew from respondents’ wave 2 reports of financial factors. Partner requests regarding money lending was measured as “In the past 12 months, has X asked you to lend him/her money?” Our measure of the respondent borrowing money from a partner was taken from the following single item: “In the past 12 months, has X lent you money?” Partner economic control/influence was measured as “Has your partner or ex-partner ever told you how to spend your money?” Financial resentment was an eight-item scale taken from a series of questions assessing respondents’ level of agreement with the following regarding finances and relationships: 1) “X made you feel bad when he/she helped you financially;” 2) You expected X to help you out financially;” 3) “X never let you forget it when he/she helped you out;” 4) “X resented having to help you out financially;” 5) “You made X feel bad when you helped him/her out financially;” 6) “X expected you to help him/her out financially;” 7) “You never let X forget when you had to help him/her financially;” and 8) “You resented having to help X out financially” (responses ranged from 1 “strongly disagree” to 5 “strongly agree”). Economic conflicts was measured as “During your relationship, how often did you and X fight about money issues?” (responses ranged from 1 “never” to 5 “very often”). A final financial measure assessed the respondents’ financial socialization based on the following single item: “Has an adult ever talked to you about good ways to choose when to save or spend your money?”

Traditional Predictors (wave 1)

A set of traditional predictors were measured at wave 1 of STRiV. Youth’s exposure to violence between parents in the home was measured by two items drawn from the National Survey of Children’s exposure to Violence (NatSCEV) (Turner, Finkelhor, & Omrod, 2010). These items were taken from the wave 1 interview, and include “At any time in your life, did you see a parent get pushed, slapped, hit, punched, or beat up by another parent or their boyfriend or girlfriend” an identical item assessed whether the respondent reported “hearing” the same description of parental violence. A positive response to one or both items was coded as a single indicator signaling witnessing parental violence (1 = yes). Respondents were asked how often in the past 12 months they had gotten into “a serious physical fight,” based on measurement in the National Longitudinal Study of Adolescent Health (Add Health). We dichotomized scores to indicate whether the respondent reported any general aggression. STRiV measured conditional tolerance for ARA, adapted from Giordano et al. (2010), based on four hypothetical situations in which respondents would “think it was OK for someone to hit” a boyfriend or, separately, a girlfriend. Response categories were a 4-point scale (strongly agree, agree, disagree, or strongly disagree). Scenarios included the boy/girlfriend making the hypothetical “someone” mad; insulting someone in front of friends; making someone jealous on purpose; cheating; or hitting the partner first. We constructed an indicator of attitudes toward partner violence representing any conditional tolerance (1) and no tolerance (0) (Mumford, Taylor, & Giordano, 2017).

Sociodemographic Characteristics

We also included controls for a range of sociodemographic characteristics including gender, age at the wave 3 interview (continuous), and race/ethnicity including white non-Hispanic (contrast category), Black, Hispanic, and other. Household income was constructed as a dichotomous variable indicating whether household earnings were above the 2013 median income (1 = yes). Parental education assesses the highest education attained in the household, coded as three indicator variables including high school or less (contrast category), at least some college education, and a college education or greater. An indicator of a two-parent household was coded as married or living together (1) or otherwise (0). Partner and respondent employment (wave 3) were measured based on the following: “In the last 4 weeks, did [you/partner] work for pay for anyone outside [your/partner’s] home?” In both cases, the questions specified that employment would reflect “both regular jobs and things like baby-sitting or yard work.” All sociodemographic characteristics (with the exception of age and employment) included in this investigation were measured at wave 1 of the STRiV survey.

Analytic Strategy

We began by presenting weighted descriptive statistics, including the mean, standard deviation, and range, for all the variables included in the current investigation. Next, multivariate analyses proceeded in two stages. In the first, we examined associations between adolescent financial factors, traditional predictors, sociodemographic characteristics, and economic conflicts. Our measure of conflict was dichotomous in nature, and thus these analyses employed logistic regression models. First, we presented zero-order associations between adolescent financial factors (wave 2), traditional predictors (wave 1), sociodemographic characteristics, and economic conflicts (wave 2). Next, we examined associations between adolescent financial factors and economic conflicts, controlling for the full range of traditional predictors and sociodemographic characteristics.

In the second set of analyses, we examined associations between adolescent financial behaviors/socialization (wave 2) and physical ARA perpetration (wave 3), controlling for the baseline traditional predictors and sociodemographic characteristics. These associations were similarly modeled using logistic regression. We began by examining zero-order associations, and proceeded to estimate a series of nested models. In the first, we entered the adolescent financial behaviors/socialization as a block. In model 2, we included controls for sociodemographic characteristics. Model 3 added the traditional predictors. To determine whether these processes differed for male and female respondents we tested gender interactions. The results of these analyses are described below.

Results

Table 1 presents the means/percentages, standard errors, and ranges of all study variables. Roughly 15% of the sample reported physical violence perpetration at the time of the wave 3 interview. Of the adolescent financial factors (wave 2), more than one-quarter (28%) of the sample reported economic conflicts. The average level of financial resentment was 1.57 (on a scale of 1 to 5), suggesting low levels of endorsement overall. Approximately one-fifth (18%) of the sample reported partner requests for money lending, and slightly fewer (14%) indicated that they had borrowed money from a partner. The experience of economic control/influence was rare, with under 7% of the sample reporting such exposure. The vast majority of respondents reported financial socialization (81%).

Table 1.

Weighted Descriptive Statistics for Study Variables (n = 379)

	Mean/Percentage	SD	Range
Dependent Variables
Physical violence perpetration (wave 3)	14.79%
Independent Variables
Adolescent Financial Factors (wave 2)  Economic conflicts	27.52%
Financial resentment	1.57	0.06	1–5
Partner requests regarding money lending	18.47%
Respondent borrowed money from partner	14.47%
Partner economic control/influence	6.69%
Financial socialization	81.15%
Traditional Predictors (wave 1)  Witnessing parental violence	21.21%
General aggression	10.61%
Attitudes toward partner violence Sociodemographic Characteristics (wave 1)a	34.08%
Gender (male)   Female	46.79%
Age (wave 3)	17.36	0.16	11–21
Race/Ethnicity (White)   Black	9.25%
Hispanic	12.51%
Other	5.53%
Partner employed (wave 3)	56.13%
Respondent employed (wave 3)	59.16%
Household Income (% > 2013 median income)	57.22%
Parental Education (high school or less)   Less than high school	6.46%
Some college	20.79%
College or more	46.95%
Two-parent household (% yes)	76.35%

^*p < .05

^**p < .01

^***p < .001

^aSociodemographic characteristics were measured at the wave 1 interview, unless otherwise indicated

With respect to the traditional predictors (wave 1), just over one-fifth of the sample reported witnessing parental violence, and roughly one in ten had gotten into a serious fight in the last 12 months. Roughly one-third (34%) of the sample endorsed the use of violence toward a romantic partner under at least one of the five conditions described.

Table 2 presents logistic regression coefficients identifying factors associated with heightened odds of conflict. At the zero order, financial resentment, partner requests regarding money lending, and partner economic control/influence were positively associated with the odds of economic conflicts. In addition, age and partner employment were associated with heightened risk, such that the odds of conflict were greater among older respondents and respondents whose partners were employed (during the month preceding the wave 3 interview). In the second column, which presents adjusted odds ratios controlling for the full range of study variables, a number of significant associations persisted. In particular, financial resentment and partner requests regarding money lending were associated with higher odds of conflict, net of other factors. Further, accounting for the full range of controls, the association between age and conflict remained positive and significant.

Table 2.

Odds Ratios for the Logistic Regression Models Predicting Conflict due to Economic Considerations (Wave 2) (n = 379)

	Zero Order	Full Model
Adolescent Financial Factors (wave 2)  Financial resentment	1.913**	2.085***
Partner requests regarding money lending	5.284***	5.252***
Respondent borrowed money from partner	2.027	0.806
Partner economic control/influence	3.940**	2.320
Financial socialization	1.255	0.688
Traditional Predictors (wave 1)  Witnessing parental violence	1.298	0.877
General aggression	0.828	1.470
Attitudes toward partner violence	1.196	0.836
Sociodemographic Characteristics (wave 1)a  Gender (male)   Female	1.733	1.636
Age (wave 3)	1.301**	1.317**
Race/Ethnicity (White)   Black	1.135	0.478
Hispanic	1.013	0.553
Other	2.233	3.129
Partner employed (wave 3)	2.243*	1.357
Respondent employed (wave 3)	1.471	0.705
Household Income	0.886	1.147
Parental Education (ref: high school or less)   Less than high school	1.172	0.621
Some college	0.638	0.474
College or more	1.142	1.036
Two-parent household	1.648	0.078
F		2.89***

^*p < .05

^**p < .01

^***p < .001

^aSociodemographic characteristics were measured at the wave 1 interview, unless otherwise indicated

Next, we estimated a series of models to predict physical violence perpetration (wave 3) based on adolescent financial factors (wave 2), traditional predictors (wave 1), and sociodemographic characteristics (Table 3). The first column presents the zero-order associations. At the bivariate level, economic conflicts and partner requests regarding money lending are positively associated with physical violence perpetration. None of the traditional predictors were associated with the odds of physical violence perpetration. In terms of the sociodemographic characteristics, gender, age, and race/ethnicity were associated with physical violence perpetration such that older respondents, respondents who are female, and Black respondents (relative to their white counterparts) experienced greater risk of physical violence perpetration. Adjusted results are presented to the right of the zero-order associations. Model 1 entered the adolescent factors as a block. Net of the other financial factors, only economic conflicts was significantly associated with the odds of physical violence perpetration. That is, respondents who reported economic conflicts at wave 2 were at greater risk of perpetrating physical violence within the context of a dating relationship at wave 3. Model 2 added the sociodemographic characteristics as a block. Net of these factors, the association between economic conflicts and physical violence perpetration remained significant and positive. After the inclusion of the sociodemographic characteristics, the association between financial resentment and physical violence perpetration became significant, suggesting that greater endorsement of such resentment at wave 2 is associated with higher odds of physical violence perpetration at wave 3. The suppression of financial resentment was due to gender. That is, both gender and financial resentment are positively correlated with IPV perpetration, however, the correlation between resentment and gender is negative. Thus, less of the variance in IPV perpetration is accounted for by financial resentment in a model in which we fail to account for gender (i.e., female). A final model (model 3) added the block of traditional predictors. Controlling for the full range of study variables, economic conflicts and financial resentment were associated with heightened odds of subsequent physical violence perpetration.

Table 3.

Odds Ratios for the Logistic Regression Models Predicting Physical Violence Perpetration (Wave 3) (n = 379)

	Zero Order	Model 1	Model 2	Model 3
Adolescent Financial Factors (wave 2)
Economic conflicts	4.572*	3.431**	3.361**	3.344**
Financial resentment	1.552	1.311	1.599*	1.613*
Partner requests regarding money lending	3.258*	2.133	1.219	1.208
Respondent borrowed money from partner	2.350	1.529	1.025	1.031
Partner economic control/influence	1.474	0.583	0.324	0.292
Financial socialization	0.729	0.544	0.553	0.566
Traditional Predictors (wave 1)
Witnessing parental violence	1.554			1.013
General aggression	0.650			1.934
Attitudes toward partner violence	1.850			1.042
Sociodemographic Characteristics (wave 1)a
Gender (male)
Female	5.386***		8.604***	9.074***
Age (wave 3)	1.226*		1.200	1.219
Race/Ethnicity (White)
Black	3.570*		8.657**	8.505**
Hispanic	0.783		0.520	0.474
Other	1.858		1.165	1.246
Partner employed (wave 3)	1.258		0.361*	0.350*
Respondent employed (wave 3)	1.163		1.033	1.026
Household Income	0.628		0.355	0.358
Parental Education (ref: high school or less)
Less than high school	1.723		4.175	4.451*
Some college	0.982		2.071	1.855
College or more	1.140		3.242	3.182
Two-parent household F	1.911	2.91**	3.726* 2.37**	3.901* 2.32***

^*p < .05

^**p < .01

^***p < .001

^aSociodemographic characteristics were measured at the wave 1 interview, unless otherwise indicated

A final set of models (Table 4) examined the potential for gender differences in the effects of adolescent financial factors on the odds of physical violence perpetration. Most of these interactions were not significant, which suggests that the effects of these particular behaviors and socialization efforts on violence are similar for male and female adolescents. There was one important exception. The cross-product term for the interaction of economic conflicts and gender was significant and positive. This suggests that the effect of economic conflicts on violence perpetration was stronger for females. Moreover, the main effect of economic conflicts, which represents the effect of conflict on violence for males, was not significant.

Table 4.

Odds Ratios for the Logistic Regression Models Predicting Physical Violence Perpetration (Wave 3), the Moderating Effect of Gender (n = 379)

Adolescent Financial Factors (wave 2)
Economic conflicts	0.445
Financial resentment	1.616*
Partner requests regarding money lending	1.196
Respondent borrowed money from partner	0.833
Partner economic control/influence	0.347
Financial socialization	0.588
Traditional Predictors (wave 1)
Witnessing parental violence	0.778
General aggression	1.618
Attitudes toward partner violence	1.041
Sociodemographic Characteristics (wave 1)a
Gender (male)
Female	3.251
Age (wave 3)	1.200
Race/Ethnicity (White)
Black	11.814**
Hispanic	0.609
Other	1.347
Partner employed (wave 3)	0.395
Respondent employed (wave 3)	1.144
Household Income	0.284
Parental Education (ref: high school or less)
Less than high school	5.527*
Some college	1.574
College or more	2.952
Two-parent household	5.207*
Economic conflicts x Female	15.664*
F	2.44***

^*p < .05

^**p < .01

^***p < .001

^aSociodemographic characteristics were measured at the wave 1 interview, unless otherwise indicated

Discussion

Prior research on ARA often draws on either adolescent or adult theories to study violence in adolescent dating relationships. In the current investigation we adopted a multidimensional framework, which included elements from the traditional theories of partner violence (e.g., social learning, feminist, and individualized approaches) in addition to more nuanced features of adolescents’ relationships. In particular, we investigated further a domain that has previously appeared almost exclusively in the marital violence literature—economic considerations. Yet we moved beyond prior work by unpacking how such considerations contribute to couple-level interactions, including conflict. Furthermore, we examined these associations longitudinally, focusing attention on the types of dynamics that contribute to economic conflicts specifically, and in turn, ARA perpetration. This is an important contribution, as prior work has largely focused on cross-sectional associations, and thus has been unable to establish temporality.

Using longitudinal data from a nationally representative sample of adolescents, we found partial support for our first hypothesis, which posited that adolescent financial factors were associated with concurrent economic conflicts. More specifically, our findings indicated that feelings of financial resentment, partners’ requests to borrow money, and partner economic control were associated with increased odds of ARA perpetration. After controlling for the complete range of study variables, the association between financial resentment and economic conflicts was attenuated; however, the links between financial resentment, requests to lend money to a partner, and conflict remained significant and positive. In addition, age was associated with elevated odds of conflict. These findings suggest that although adolescents may not be financially interdependent in their dating relationships to the same extent as adults, they do face certain financial decisions, experience economic control, and develop feelings of resentment on the basis of financial factors. Furthermore, the experience of being asked for money from a partner and having financial resentment are associated with increased conflict within the relationship. That older respondents are at greater risk of experiencing these economically motivated conflicts likely reflects their involvement in more serious, longer-term relationships. It may also reflect the fact that older youth are beginning to acquire more financial responsibility as they make moves toward establishing economic independence from parents.

In analyses predicting future ARA perpetration, we found partial support for hypothesis 2. In particular, we found that conflict due to economic considerations was an important predictor of future ARA perpetration. This is consistent with prior cross-sectional work that identified an association between economic concerns as a source of conflict and IPV among a sample of young adults (Copp, Giordano, Manning, & Longmore, 2016), and further emphasizes the need for additional research to look beyond more general measures of verbal conflict to consider the specific stressors (e.g., financial considerations/decision-making) that lead to relationship conflict. Relatedly, financial resentment was also associated with heightened odds of ARA perpetration, suggesting that couple-level resentment regarding financial matters may represent an area of conflict in adolescent relationships. Results also indicated that partner requests for money were associated with ARA perpetration at the zero order, however, this association was explained after controlling for the other financial factors. Supplemental analyses revealed that this attenuation was driven by the addition of economic conflicts to the model. Taken together with the findings from our analyses examining predictors of economic conflict, this pattern suggests that partner requests for money contribute to couple-level conflict about financial issues, which in turn increases the odds of violence perpetration.

Consistent with prior work which has identified the potentially gendered nature of the mechanisms driving the use of violence within the dyadic context, we also considered whether these associations were similar for male and female adolescents. Analyses in support of hypothesis 3 found that the association between economic conflict and ARA perpetration was stronger for female adolescents. Moreover, supplemental analyses revealed that the link between economic conflicts and ARA perpetration was not significant for the young males included in this investigation. It may be the case that female adolescents are more likely to perceive their partners as financially irresponsible, or they may feel that they contribute less and/or are less invested in the relationship, thus fueling conflict. Alternatively, stereotypical expectations of male economic contributions, and disappointment surrounding reality, may fuel these conflicts. However, it is unclear why economic conflict is more closely linked to the use of violence among female adolescents. Future research is needed to more fully describe the inner-workings of adolescent relationships, including the types of conflict that escalate to the point of violence.

In addition to our focus on proximal relationship/situational factors, this investigation drew on prior research and theorizing in the social learning, feminist, and psychological traditions to further round out our understanding of the etiology of ARA. Yet surprisingly, we did not find much support for social learning, feminist, or individualized approaches in our models. On the one hand, it is unsurprising that the more proximal features of the relationship would matter more for IPV perpetration. However, that these theoretical constructs were not significantly associated with ARA perpetration at the zero order suggests that future work should focus additional empirical attention on the role of social learning, feminist, and psychological mechanisms in adolescent relationships characterized by violence.

Although this study offers new contributions, there are some limitations. First, similar to other research on ARA based on self-report surveys, the measure of perpetration used in the current investigation is taken from the CADRI. Thus, the current study results must be interpreted within the limitations inherent in the CADRI, including a lack of attention to the context surrounding the violent event and motives for the resort to violence. Second, although the STRiV data are nationally representative, there were a number of necessary sample exclusion criteria that may lessen our ability to generalize the findings to all adolescents in the United States. In particular, attrition analyses revealed that those retained in the longitudinal sample did differ slightly from the baseline sample (see methods section for more details); however, sample selection models suggested that selection bias was not a concern in our data. Third, the unit of analysis of the current investigation is the individual as opposed to the couple. Accordingly, the items assessing the partners’ requests relied on respondents’ reports. Future research may benefit from couple-level data to provide a more dyadic approach to the relationship dynamics that foster violence. Fourth, this paper focused on associations between economic considerations and ARA for the average adolescent; however, there is the potential for heterogeneity in the observed effects. In particular, there may be couple-level variation in the trajectories of ARA on the basis of economic considerations and other relationship dynamics. Finally, although the sample is nationally representative, the majority of respondents resided in household with earnings above the median income. It is likely that financial socialization and financial considerations in the dyadic context differ according to socioeconomic status. Future research should direct additional attention to the role of economic factors in ARA using samples that include a greater share of respondents from lower income levels.

Despite these limitations, the findings from the current investigation lend additional support to a relational view of partner violence, and suggest the need for more comprehensive prevention programming. Such programming may include essential components of financial literacy and gender differences in financial risk tolerance as well as discussions of financial conflict—all areas not typically addressed in ARA interventions. Furthermore, our results suggest the potential utility of linking broader financial management training for youth with ARA prevention programming. Indeed, our findings revealed that the vast majority of youth had spoken to an adult about money-spending habits. However, these conversations are perhaps unstructured and did not appear to influence the likelihood of fights in the relationships due to financial factors or the use of relationship violence. Accordingly, programmatic and educational efforts that target these potential economic conflicts within the relationship may be more beneficial for the development and maintenance of healthy relationship behaviors. This type of programming does not have to be designed as a new intervention. A number of the evidence-based ARA prevention programs such as Safe Dates (Foshee et al., 1998) and Shifting Boundaries (Taylor et al., 2013) could add financial management and conflict components to their curricula. The programming would educate youth on the link between financial management and financial conflicts and ARA and how to interrupt this cycle. Parent-based interventions can also be designed to go beyond broad messages about responsible money management to include more specific communication on how parents can teach their teen how to negotiate financial matters in their teen relationships and address financial conflict to prevent ARA. Although this study did document a link between economic conflicts and future IPV perpetration, there is a need for more long-term longitudinal research on whether experiences in adolescent romantic relationships lead to relationship behaviors that carry forward into adulthood—including how youth negotiate financial decisions in their early dating relationships.