State Laws Addressing Teen Dating Violence in US High Schools: A Difference-in-Differences Study

Abstract

Objective:

Teen dating violence (TDV) is prevalent with lifelong adverse consequences, and strategies to reduce its burden are needed. Many U.S. states have enacted laws to address TDV in schools, but few studies have examined their effectiveness. This study aimed to assess whether state TDV laws were associated with changes in physical TDV victimization among high school students.

Methods:

We used repeated cross-sectional data of high school students from the Youth Risk Behavior Survey across 41 states from 1999 to 2019. Using a difference-in-differences approach with an event study design, we compared changes in past-year physical TDV in states that enacted TDV laws (n=21) compared to states with no required laws (n=20). Analyses accounted for clustering at the state-level and state and year-fixed effects. We conducted sensitivity analyses to assess the robustness of our findings.

Results:

In our sample of 1,240,211 students, the prevalence of past-year physical TDV was 9.2% across all state-years. In 1999, the prevalence of TDV at the state-level ranged from 7.5–13.0%; in 2019, the prevalence ranged from 3.7–10.5%. There was no significant association between TDV laws and past-year physical TDV. Six or more waves after enactment, we observed a non-significant 1.7% percentage point reduction in TDV in states with TDV laws (95% CI: −3.6 to 0.3 percentage points; p=0.10).

Conclusions:

We found no significant association between enactment of TDV laws and physical TDV among high school students. Further research is needed to understand how TDV laws are implemented and components of TDV laws that may influence effectiveness.

Introduction

Teen dating violence (TDV) remains a substantial public health problem for adolescents, families, and communities [1]. Approximately one in eight U.S. high school students experience physical and/or sexual violence by someone they are dating each year [2]. One nationally-representative survey in 2013–2014 of 12 to 18-year-old dating youth found that more than 60% experienced psychological/emotional abuse [3]. These experiences of TDV can have negative physical health, mental health, and social consequences, including depression, anxiety, suicidal ideation, increased substance use, decreased academic achievement, subsequent intimate partner violence, and in extreme cases, death [4–6]. Although TDV is widespread with lifelong adverse consequences, it is preventable, and strategies to reduce the burden of TDV are needed [7,8].

Schools provide an important, structured setting for TDV prevention and response, by creating caring climates, providing support for student well-being, and implementing prevention programs [9]. Over the past couple decades, states have increasingly recognized the seriousness of TDV, the central role schools can play, and have enacted laws requiring schools to address TDV. These laws are structural interventions that lay a foundation for a range of activities in schools, including education for students and training for school staff [10–12]. They can facilitate scaling up the use of multilevel strategies, including implementing educational programs that have been shown to reduce violence but are often not widely used [13–17]. Importantly, these laws respond to calls to prioritize societal strategies that affect the outer layers of the social ecology to increase population-level impact [13,18].

As of September 2020, 29 states had at least 1 law related to TDV requirements in schools, with variability in the requirements across states [14]. These laws are often similar to anti-bullying laws, which have been shown to be effective in reducing school bullying [10,19,20]. Both TDV and anti-bullying laws can include, for example, definitions of the behaviors, prevention education requirements, and requirements for districts to have written policies and detailed response procedures [14,21]. States with TDV laws all require schools to provide prevention education to students, most commonly including content on healthy relationships and awareness-raising of TDV [14]. Despite the proliferation of TDV laws, significant gaps remain in understanding their effectiveness. To our knowledge, only one prior cross-sectional study examined these laws and found no relationship between the presence of a TDV law and TDV prevalence among high school students [22]. We build on the limited existing literature by capitalizing on between-state variation in the enactment of TDV laws paired with repeated cross-sectional data to estimate the association between TDV laws and TDV victimization using a difference-in-differences approach. While we are unable to measure implementation of the laws, we assess whether the enactment of state TDV laws is associated with changes in physical TDV victimization among high school students. These results can inform policymakers and assess whether enacting laws may serve as an effective population-level prevention strategy for TDV.

Methods

We obtained repeated cross-sectional data from the Youth Risk Behavior Survey (YRBS), a biennial survey of health behaviors of 9^th through 12^th grade students conducted by the Centers for Disease Control and Prevention. States employ a two-stage, cluster sample design to obtain state-level representative samples of high school students. Students anonymously and voluntarily complete self-administered questionnaires. Weights are adjusted for student nonresponse and distribution of students by grade, sex, and race/ethnicity for states with a response rate ≥60%. We used all years of available data for 41 participating states from 1999–2019 with a response rates of ≥60% and data-sharing agreements with the CDC [23]. The University of Washington Institutional Review Board did not require approval for this study given the use of publicly available, deidentified data.

Exposure

The exposure of interest was enactment of state laws that require schools to address prevention and/or response to TDV (e.g., required prevention programming, required school districts to develop a written policy for TDV), gathered through longitudinal policy surveillance. Comprehensive details on the content of these laws, including variation and state-specific data, have been published previously [14]. There were 21 states with pre- and post-law YRBS data that enacted a required TDV law during the study period (1999–2019): Alaska, Arkansas, Connecticut, Delaware, Florida, Idaho, Illinois, Iowa, Kentucky, Louisiana, Michigan, Nebraska, New Hampshire, New Jersey, New Mexico, North Carolina, Oklahoma, Rhode Island, Tennessee, Vermont, and West Virginia (Supplementary Table 1). Analyses used individual-level data, where each YRBS participant was classified as exposed if they lived in a state where at least one component of a TDV law was required (treatment group; n=21 states). Participants who were exposed were used to estimate the treatment effect after enactment of the law if the law was in effect for at least 6 months of the survey year, since YRBS is administered biennially in the spring and laws take time to implement and take effect. Participants living in a state without a law or with a law that was solely ‘encouraged’ (i.e., laws encourage but don’t require that schools address TDV) were classified as unexposed (control group; n=20 states). We excluded three states (California, Texas, Virginia) that lacked YRBS data before and/or after enactment of the law.

Outcome

Our outcome of interest was past-year physical TDV. From 1999–2011, this was assessed with the question: “During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose?” with a binary yes/no response option. From 2013–2019, a revised question was used: “During the past 12 months, how many times did someone you were dating or going out with physically hurt you on purpose? (Count such things as being hit, slammed into something, or injured with an object or weapon).” Response options were: “I did not date or go out with anyone during the past 12 months”, “0 times”, “1 time”, “2 or 3 times”, “4 or 5 times”, and “6 or more times”. To align responses from both time periods, we included all students in the denominator and dichotomized the measure to no/0 vs. yes/≥1 times (i.e., students who indicated they did not date or go out with anyone were coded as no/0). Students missing outcome data were excluded (4.8% of students across all states in 1999–2019). In sensitivity analyses including only the years 2013–2019, we used the “I did not date or go out with anyone during the past 12 months” response option to exclude students who did not date from the denominator, given that they were not at risk of experiencing TDV. This calculation of prevalence is in line with the CDC’s current estimates [2].

Covariates

In adjusted analyses, we controlled for a set of time-varying state-level covariates that a priori could have changed over the study period differentially by state and are plausibly related to the exposure and outcome. These included percent low-income uninsured children, National School Lunch program participation, and political party affiliation (i.e., whether the Governor was a Democrat), all available from the University of Kentucky Center for Poverty Research [24]. We also included state pupil-teacher ratio from the U.S. Department of Education as an indicator of school resources and enactment of anti-bullying laws given the overlap between these and TDV laws [25,20]. Covariates were available for all state-years of the study period. Our main analyses did not control for individual-level demographics since they had no a priori association with TDV laws and are not considered confounders; this lack of association was confirmed in our data.

Main analyses

We first summarized demographic characteristics for students overall and in states that did and did not enact a TDV law during the study period. We examined the prevalence of TDV over time by states that did and did not enact a TDV law. We then estimated the association between TDV laws and changes in TDV using a difference-in-differences (DiD) approach. This approach compared TDV among students who were exposed to a TDV law to students who were not exposed to a TDV law. In our main analysis, we estimated an average treatment effect after enactment of a law using a binary variable equal to one for students living in a state after the state’s TDV law was enacted and a zero for students living in a state before TDV law enactment or in a state that never enacted a TDV law.

To assess the validity of the DiD approach in this context, we examined the prevalence of TDV in each state over time comparing states that did and did not enact a law during the study period. Importantly, we estimated an “event-study” specification where differences in outcomes between states implementing a TDV law versus those that did not were allowed to vary over time [26]. A key assumption of the DiD approach is the parallel trends assumption—that is, outcomes in states with TDV laws would have continued along their same trajectories in the absence of a law. While the assumption cannot be directly verified, the event study analyses can be used to assess potential violations by examining potential differences between exposed and unexposed students in years prior to enactment of a law. The event-study approach also permits the assessment of time-varying associations denoting leads and lags. For time-varying associations, we present estimates for six or more waves pre-law and six or more waves post-law, given the low number of states contributing data to those estimates several waves before and after the enactment of the laws. The reference period (t=0) refers to the survey wave before TDV law enactment (i.e., prior to TDV laws being in effect for at least six months).

As TDV laws were enacted in different years during the study period, we used a staggered DiD estimator [27]. We report both an average effect (i.e., a single coefficient summarizing the average estimates across all post-law waves) in addition to the time-varying associations. Unadjusted analyses included state and year fixed-effects to account for time-invariant state-level confounders and secular trends. Adjusted analyses additionally included five time-varying state-level covariates detailed above. All analyses were conducted in Stata 15.1 (StataCorp), with standard errors clustered at the state-level (i.e., the level of law enactment) and weights to account for the complex sampling design [23]. Statistical significance was determined at the level of p<0.05.

Sensitivity analyses

For sensitivity analyses, we first estimated results additionally controlling for individual-level demographics (age, grade, sex, and race/ethnicity). We then estimated results using standard two-way fixed-effect regressions rather than the staggered DiD estimator, using both linear and logistic regression for the binary TDV outcome since estimation of treatment effects can be sensitive to functional form in the DiD framework and the TDV outcome is binary [28]. Next, we used an alternative exposure definition where participants were considered exposed if the law was in effect for at least 12 months prior to the survey year (rather than at least 6 months described above) since participants were asked about past-year TDV experiences. Lastly, we estimated results using the staggered DiD estimator with data only from 2013–2019 so we could calculate TDV prevalence among students who reported dating in the last 12 months (i.e., were at risk of experiencing TDV). For this analysis, there were 7 states with a TDV law with available pre- and post-law data between 2013–2019: Alaska, Arkansas, Illinois, New Hampshire, North Carolina, Oklahoma, and West Virginia. States with TDV laws that had only post-law data in this period were excluded since no pre-law data was available to assess potential parallel trends violations.

Results

From 1999–2019, there were a total of 1,240,211 students across 41 states in our sample. Among these students, the average age was 16.0 years with a fairly even split across 9^th to 12^th grades (Table 1). Overall, 50.6% of students identified as male, 61.7% identified as White, 18.3% as Black or African American, 13.1% as Hispanic/Latino, and <3% as American Indian/Alaska Native, Asian, Native Hawaiian/Pacific Islander, and multiple races (non-Hispanic). Demographic characteristics were generally similar between students in states with and without TDV laws. The average prevalence of past-year physical TDV among all students was 9.2% across all state-years. The prevalence of TDV declined from 1999–2019 with similar trends over time in states with and without TDV laws (Figure 1). This decline was observed to differing degrees across states (Supplementary Figure 1). In 1999, the prevalence of TDV at the state-level ranged from 7.5% (Vermont) to 13.0% (Mississippi); in 2019, the prevalence ranged from 3.7% (Iowa) to 10.5% (Louisiana).

Table 1.

Sample characteristics for Youth Risk Behavior Survey (YRBS) study participants, 1999–2019

	All students (n = 1,240,211)	States with TDV laws (n = 578,313, 51.2%)	States without TDV laws (n = 661,898, 48.8%)
Demographic	n (%)
Age, years
Mean (SD)	16.0 (1.3)	16.0 (1.2)	16.0 (1.3)
Grade
9th	345,460 (27.9)	161,274 (27.9)	184,186 (28.0)
10th	328,939 (25.8)	152,274 (25.8)	176,215 (25.8)
11th	297,989 (23.6)	138,676 (23.8)	159,313 (23.4)
12th	43,498 (22.3)	109,670 (22.2)	133,828 (22.4)
Ungraded or Other Grade	10,439 (0.3)	9,190 (0.3)	1,249 (0.3)
Sex
Male	603,887 (50.6)	282,610 (50.6)	321,277 (50.6)
Female	627,861 (49.4)	291,384 (49.4)	336,477 (49.4)
Race/ethnicity
American Indian/Alaska Native	28,482 (1.5)	14,428 (1.8)	14,054 (1.3)
Asian	49,504 (2.5)	15,340 (1.8)	34,164 (3.2)
Black or African American	156,744 (18.3)	55,833 (17.6)	100,911 (19.1)
Hispanic/Latino	188,348 (13.1)	88,656 (13.5)	99,692 (12.6)
Native Hawaiian/Pacific Islander	16,428 (0.5)	3,353 (0.4)	13,075 (0.7)
White	714,815 (61.7)	365,120 (62.5)	349,695 (60.8)
Multiple races (non-Hispanic)	56,203 (2.3)	23,492 (2.4)	32,711 (2.3)
Outcomes
Physical TDV (1999–2019)	103,324 (9.2)	47,021 (9.2)	56,303 (9.3)
Physical TDV (only 2013–2019)*	30,596 (9.3)	5,775 (9.3)	24,821 (9.3)

Note: We used data from 41 YRBS states with weighted data. Ns are unweighted, % and means are computed using sample weights to be representative of state population of high school students by student sex, grade, and race/ethnicity. Age, grade, sex, and race/ethnicity are based on participant self-report.

Missing data: age (n=3,723), grade (n=13,886), sex (n=8,463), race (n=29,687)

^*For 2013–2019, physical TDV is calculated only among students who reported dating in the past 12 months (this response option was added in 2013) and including the smaller subset of states used for 2013–2019 sensitivity analysis. The denominator for physical TDV (2013–2019) is n=308,997.

Figure 1.

Mean prevalence of physical TDV over study period across US states enacting and not enacting TDV laws, 1999–2019

Note: Prevalence of physical TDV is calculated among all students whether or not they dated in the past 12 months. From 2013–2019, YRBS revised the question to assess physical TDV and included a response option of “I did not date or go out with anyone during the past 12 months.” We observed that the percent of students who chose this option increased over time (2013: 28%, 2015: 33%; 2017: 35%; 2019: 37%). Among students who reported dating or going out with someone, prevalence of physical TDV also decreased (2013: 10%, 2015: 9%; 2017: 9%; 2019: 8%). Thus, the decrease observed in physical TDV reflects both fewer students dating and decreased reporting of TDV within dating relationships.

The event study specification suggested similar pre-enactment trends in TDV among high school students in states who enacted TDV laws compared to students in states that did not (Figure 2). However, two waves prior to TDV law enactment, there was a significant difference in TDV among students in states enacting TDV laws compared to students in states that did not (0.6% percentage point difference, 95% CI: 0.2 to 1.1 percentage points; p=0.01).

Figure 2.

Event study: TDV law enactment and physical TDV, 1999–2019

Note: Each dot plots coefficient estimates from the event study specification and vertical bars denote 95% CIs. The x-axis refers to the time (in waves) relative to implementation of TDV laws, and the y-axis is the estimated percentage point difference in TDV between states that enacted a TDV law and states that did not. Estimates were obtained using a staggered implementation difference-in-differences design, and standard errors were clustered at the state-level. The regression was adjusted for state-level covariates: percent low-income uninsured children, National School Lunch program participation, governor political party affiliation, pupil-teacher ratio, and anti-bullying laws.

In the main analysis, there was no statistically significant difference, on average, in TDV following enactment of TDV laws, accounting for the staggered enactment of TDV laws, state and year fixed-effects, and state-level covariates (Table 2). In the event study specification, we were unable to reject the null hypothesis of no differences in TDV between states that enacted laws and those that did not in any post-enactment period (Figure 2). We did observe a non-significant increase in TDV two to five waves post-enactment for students in states with TDV laws. In addition, six or more waves after enactment, we observed a reduction in TDV in states with TDV laws; however, this 1.7% percentage point reduction was not statistically significant (95% CI: −3.6 to 0.3 percentage points; p=0.10).

Table 2.

Average difference-in-differences estimates for the association between state TDV laws and prevalence of TDV across all post-law waves, 1999–2019

	Physical TDV (1999–2019)
	Percentage points (95% CI)
	Unadjusted	Adjusted
Post-TDV law enactment	−0.19 (−0.73, 0.35) p = 0.49	0.06 (−0.61, 0.73) p = 0.86

Note: Estimates are from linear regression using a staggered implementation difference-in-differences design. Unadjusted regressions control for state and year fixed effects. Adjusted models additionally include state-level covariates: percent low-income uninsured children, National School Lunch program participation, governor political party affiliation, pupil-teacher ratio, and anti-bullying laws

Findings were similar when additionally adjusted for individual-level demographics (Supplementary Table 2). These findings were robust to sensitivity analyses using standard two-way fixed effects regressions with both linear probability and logistic functional form (Supplementary Table 3, Supplementary Figure 2). In addition, findings were similar when participants were considered exposed if the TDV law was in effect for at least 12 months prior to the survey year (Supplementary Table 4, Supplementary Figure 3). When restricting the study period to 2013–2019, we found substantively similar null results for the average association between TDV laws and TDV (Supplementary Table 5). In the event study specification, TDV laws were significantly associated with a 0.9% percentage point reduction in TDV three waves after enactment (95% CI: −1.4 to −0.4 percentage points; p<0.001) (Supplementary Figure 4). Importantly, this significant difference three waves after enactment is driven by comparing students in Illinois to students in states without TDV laws.

Discussion

In this national study conducted using data from 1999 to 2019, we examined the association between the enactment of state-level TDV laws and experiences of past-year TDV victimization. We found no significant association between TDV laws and past-year physical TDV experiences among state-level representative samples of high school students in 41 U.S. states. These results align with a prior cross-sectional study using 2015 YRBS data which found no association between the presence of a TDV law and TDV outcomes [22]. Importantly, our study estimates this null finding using a rigorous, repeated cross-sectional study, employing a difference-in-differences approach to examine changes in TDV before and after the enactment of laws.

There are several plausible reasons for these null findings. The first is that while TDV laws have been enacted in certain states, they may not be widely or consistently implemented, reducing their effectiveness. We were unable to measure implementation of the laws, simply the enactment or presence of the written statutes. Research on state laws related to other interpersonal violence issues, such as bullying and suicide prevention, highlights substantial gaps between laws and policies and school practices and underscores the importance of implementation in determining effectiveness of state laws [29–32]. Low awareness of laws and limited resources have been identified as barriers of implementing anti-bullying laws [33]. Given the content of TDV laws,[14] implementation necessarily requires resources – such as staff time and funding – which are often in short supply in schools and only sometimes accounted for in the text of the laws. Limiting existing research on implementing TDV laws in schools suggests variable implementation and a need for clarity on how to implement these statutes [34,35]. In addition, even while some states specify explicit consequences for noncompliance with the law, the degree of enforcement is unknown. Further research is needed to determine if and how implementation of TDV laws is occurring. Second, TDV laws may increase awareness of TDV, creating a culture where students are more likely to disclose victimization on a survey. In that case, increased identification and disclosure of TDV could be seen a benefit of the law, which may ultimately lead to greater prevention and response efforts. Our event study specification showed a non-significant increase in TDV for two to five waves post-enactment, indicating that this may be occurring. Finally, it is also conceivable that TDV laws take substantial time to have any measurable effect on TDV outcomes. The laws are complex with multiple components, sometimes requiring state departments of education to first create models that then get implemented in districts and individual schools. Our event study specification showed a non-significant decrease in in the likelihood of experiencing TDV six or more waves post-enactment, potentially pointing to this explanation. Finally, our null findings may be a result of being underpowered to detect a statistically significant association between enactment of TDV laws (a state-level exposure) and self-reported experiences of TDV. Continued investigation into the effectiveness of these laws is warranted.

School-based prevention and response is only one avenue for addressing TDV. Even in states with robust school-based TDV laws, other services are needed to comprehensively prevent and respond to TDV, including access to medical and reproductive care and ability to obtain civil protection orders [36,37]. Outside of schools, physicians also play an important role in addressing TDV, as pediatric primary care settings offer an opportunity to engage youth and families [38,39]. Programs that engage parents have also been found to be promising for reducing TDV [40,41]. In addition to directly influencing how schools address TDV, state laws can also help establish societal norms and a culture that treats TDV seriously. While schools are a feasible, structured setting for reaching youth, comprehensive response to TDV will require multi-sectoral and multi-level approaches.

Our findings have several limitations. First, differences in TDV two waves pre-enactment for states with and without laws indicate that states potentially enacted TDV laws in response to higher levels of TDV. While we observed this difference only in a single wave, this may undermine our ability to draw causal conclusions based on our analytic approach. Additionally, TDV laws often have multiple components, such as requirements for school districts to develop written TDV policies or designating an individual to coordinate TDV efforts [14,42,43]. Our analysis did not account for comprehensiveness or components of these laws, which can vary across states. The laws we examined do, however, share characteristics such as requiring prevention education for students. In our sensitivity analyses for 2013–2019 data, the estimate based on students in Illinois showed a significant reduction in TDV post-enactment, which may indicate specific effective components of the Illinois law. Future analyses using the revised TDV survey question would be helpful given the potential lack of statistical power for the current 2013–2019 analysis. Future research on the effectiveness of specific components of TDV laws would also be useful for informing improvements in these laws.

Importantly, our analysis was limited by data availability in YRBS, which is not conducted in all states and only biennially. Our main outcome, physical TDV, is based on self-report, which is the standard approach to assessing TDV given that most youth do not report formally [44]. This outcome was based on a single question from the YRBS that asked only about physical TDV, and the survey wording changed in 2013. Since this change occurred across all states, it was accounted for using year fixed effects and did not affect the findings. TDV captured from 1999–2011 was more limited in scope (e.g., only physical abuse and only related to a boyfriend or girlfriend rather than someone youth were dating or going out with) and did not ask whether youth had dated. TDV also importantly manifests in other ways such as psychological/emotional and sexual abuse in addition to cyber abuse [3,45]. While sexual TDV was included in YRBS from 2013–2019, more states opted to exclude this question compared to the physical TDV question, limiting our ability to examine this outcome in our analytic framework. Representative state-level data with more granular measures of TDV would be beneficial for analyses of law and policy effectiveness, which are sorely needed for informing structural and population-level interventions for addressing TDV.

Conclusion

State laws addressing TDV in schools can increase awareness of TDV, allocate resources to preventing and responding to TDV, and change school culture and practices. However, our study was not able to detect significant reductions in physical TDV as a result of enacting these TDV laws. Laws are continually evolving, providing opportunities to strengthen the comprehensiveness and specificity that could potentially increase their effectiveness in reducing TDV.

Highlights.

• We examined effectiveness of enacting state laws addressing teen dating violence in high schools

• We found no association between teen dating violence laws and past-year physical dating violence

• Non-significant trends indicated reduced dating violence after several waves of law enactment

• Future studies with additional years of post-enactment data are warranted

Abbreviations:

TDV

teen dating violence

YRBS

Youth Risk Behavior Survey

DiD

difference-in-differencess