Abstract
We examine the heterogeneous effects of childhood bullying victimization and chronic bullying victimization on a wide range of outcomes using data from the National Longitudinal Survey of Youth in the United States. Bullying victimization is categorized as childhood or chronic depending on the duration of victimization. We find that bullying victimization negatively affects the physical and mental health of youths, and increases the likelihood of engaging in risky behaviors, such as, smoking and marijuana use. The adverse effects tend to increase in magnitude with chronic bullying victimization. In addition, chronic bullying victimization increases the likelihood of utilizing mental health services and experiencing life-disrupting emotional problems in adulthood.
Supplementary Information
The online version contains supplementary material available at 10.1007/s40653-023-00542-0.
Keywords: Bullying victimization, Mental health, Risky behaviors
Introduction
Bullying is defined as repeated direct or indirect acts of aggression against or towards a person, accompanied by power imbalance (Olweus, 1993). In the United States, about 22% of six- to eleven-year-old children (Lebrun-Harris et al., 2020) and 20% of twelve- to eighteen-year-olds (National Center for Education Statistics, 2020) have been victims of bullying. Increasing evidence shows that being a victim of bullying in childhood negatively affects a victim’s well-being and contributes to a plethora of mental health problems that can persist into adulthood (Singham et al., 2017; Takizawa et al., 2014). Children and adolescents who were exposed to bullying are at higher risk of substance use, suicidal ideation (Quinn & Stewart, 2018), and suicidal attempts (Baiden & Tadeo, 2020; Geoffroy et al., 2016). Childhood bullying victimization is associated with an increased likelihood of using mental health services in childhood, adolescence, and in midlife (Evans-Lacko et al., 2017). In addition to the direct health concerns for those who are bullied, significant and negative economic consequences are associated with bullying victimization. Bullying victimization has been found to be negatively associated with educational performance (Oliveira et al., 2018; Ponzo, 2013; Brimblecombe et al., 2018) found significant negative impact on lifetime wealth and employment opportunities among those who experienced childhood bullying.
This research assesses the extent to which there are differences in effects of childhood and chronic bullying victimization with respect to health outcomes, educational attainment, and utilization of mental health services. The results suggest that both childhood and chronic bullying victimization are associated with poorer mental and physical health. Importantly, in comparing impacts of childhood and chronic bullying victimization, mental health outcomes are worse among individuals who were exposed to chronic bullying compared to those who experienced bullying during early childhood but not in adolescence. Chronic bullying victimization is also found to be associated with significantly greater utilization of mental health services in adulthood.
An understanding of the effects of childhood and chronic bullying victimization has important policy implications. It is important to start anti-bullying interventions as early as possible since youth who were bullied repeatedly over longer duration have worse mental health outcomes. Many victims do not seek timely therapy from mental health professionals because of the perceived stigma of mental treatment, financial, cultural, or other reasons (Mojaverian et al., 2013; Mojtabai, 2005). In comparing attitudes toward seeking mental health services between Japanese and American undergraduate students, Mojaverian et al. (2013) found cultural differences in seeking professional help and that the seeking behaviors could be partially mediated by use of social support. Mojtabai (2005) found significant cost barriers among uninsured individuals in need of mental health care in the United States. Investing in anti-bullying programs and improvements in health insurance coverage could help alleviate the long-term impacts on mental health and productivity associated with bullying victimization.
Literature Review
Various studies have found that bullying victimization is associated with increased risk of mental health problems (Brendgen & Poulin, 2018; Lee & Vaillancourt, 2018; Lereya et al., 2015; Schnyder et al., 2017; Sigurdson et al., 2015; Stewart-Tufescu et al., 2021), suicide attempts, and suicidal ideation (Baiden & Tadeo, 2020; Geoffroy et al., 2016). The negative impacts on academic performance (Eriksen et al., 2014; Oliveira et al., 2018; Ponzo, 2013), on educational attainment and wages in adulthood (Brown & Taylor, 2008; Connolly et al., 2019; Fernandez et al., 2015), and on risky behaviors (Boynton-Jarrett et al., 2013; Bouffard & Koeppel, 2014; Sarzosa & Urzúa, 2021) are well documented.
There are also repeated and consistent findings of longer-term impacts of bullying victimization in adulthood. For instance, Takizawa et al. (2014) showed that the victims of childhood bullying are more likely to have depression and anxiety disorders in middle age, poor general health across a wide age range, lower educational attainment in midlife, poor cognitive functioning, and poor self-perceived quality of life at age fifty. Brunstein et al. (2019) analyzed longitudinal surveys of adolescents from ten European countries and found that chronic victimization is associated with an increased likelihood of experiencing depression, suicidal ideation, and suicide attempts relative to sporadic victimization. Using data from the Rural Adaptation Project in North Carolina, Evans et al. (2019) showed that youth who experienced bullying victimization for over five years in middle school and high school had higher levels of depression and lower levels of self-esteem and future optimism. Hoffman et al. (2017) compared effects of childhood, adolescent, and chronic bullying victimization on later life outcomes. They found that childhood victimization and chronic victimization were positively associated with depression, substance use, and violence in adulthood. The current paper adds to this rich literature on bullying victimization by examining the heterogeneous effects of experiencing childhood bullying and chronic bullying on self-reported health status, risky behaviors, and use of mental health services in a nationally representative panel data from the United States.
Data
In this study, panel data from the National Longitudinal Study of Youth 1997 (NLSY97) cohort are used. The first round of the NLSY97 surveyed a nationally representative sample of 8,984 youth who were between twelve and sixteen years of age as of December 31, 1996. Subsequent and ongoing surveys have tracked the transition of young individuals living in the United States from school into adulthood. The annual NLSY97 surveys include extensive information on family background, childhood, educational experiences, health, and employment from a cross-sectional sample of 6,748 youth and an over-sample of 2,236 Hispanic and non-Hispanic black youth born between 1980 and 1984. This study uses data from the 1997 to 2009 rounds of NLSY97. Specifically, health outcomes and substance use indicators measured in the 2004 wave, when participants were between nineteen and twenty-five years old, are considered. Data regarding the use of mental health services and any experience of life-disrupting problems in adulthood are drawn from the 2009 survey year when these questions were first administered and when participants were between twenty-four and thirty years old. The response rates remained above 80% across the waves.
Victimization Measures
NLSY97 measures bullying victimization events before the age of twelve years and between the ages of twelve and eighteen years. During the 1997 survey, respondents were asked if they had experienced repeated bullying before the age of twelve years. After respondents turned eighteen years of age, they were asked if they were ever victims of repeated bullying between the ages of twelve and eighteen years. Responses to these two questions about bullying experiences are transformed into two measures of bullying victimization: childhood and chronic bullying victimization. The first measure, childhood bullying victimization, is assigned a value ‘1’ if the respondent reported being bullied before the age of twelve years (and were not bullied in later years) and it is ‘0’ if a respondent had no history of being bullied. The second measure, chronic bullying victimization, is assigned the value ‘1’ if a respondent was a victim of bullying before the age of twelve years and between twelve and eighteen years, and it is ‘0’ if they did not experience any bullying. A small percentage of individuals had reported that they had not experienced childhood bullying, but that they had experienced bullying in later adolescence. They are excluded from subsequent analysis due to the small sample size. Descriptive statistics of the two victimization variables are presented in the first two rows of data in Table 1. In the sample of 5,818 individuals, 16.5% reported childhood bullying victimization, while 5.4% individuals in the sample of 5,136 individuals reported chronic victimization.
Table 1.
Descriptive statistics of victimization and outcome variables
| Outcome variables | Frequency or Mean (SD) |
|---|---|
| Childhood bullying victimization (n = 5,818) | 16.50% |
| Chronic bullying victimization (n = 5,136) | 5.41% |
| Mental health score† (n = 5,480) | 15.55 (2.43) |
| Good general health (n = 5,695) | 93.13% |
| Drinking (n = 5,641) | 73.78% |
| Smoking (n = 5,672) | 42.72% |
| Marijuana use (n = 5,604) | 21.29% |
| High School Diploma (n = 6,393) | 71.50% |
| College attendance (n = 4,571) | 51.00% |
| Age at first employment, years (n = 6,149) | 20.84 (3.15) |
| Use of mental health service (n = 5,533) | 6.24% |
| Life-disrupting emotional/mental problems (n = 5,529) | 5.39% |
† The mental health score is measured on a Likert scale, ranging from 5 to 20 where a higher value indicates a better mental health status
Notes. All data, except the last two variables, are from the 2004 survey of the NLSY97. Data for the last two variables are from the 2009 survey. Mental health score is a discrete variable and age at first employment is a continuous variable. Remaining variables are 0–1 indicator variables
Outcome Variables
There are ten outcome variables. They include self-reported mental health measure, general health status, three substance use variables, high school diploma attainment, college attendance, timing of first employment, utilization of mental health services, and any experience of life-disrupting mental or emotional problems. The mental health measure is based on responses to five questions regarding how often during the previous month they had felt calm or peaceful, happy, nervous, down or blue, and depressed. Each indicator was assessed on a 4-point Likert scale (‘all of the time’ was scored as ‘1’, ‘most of the time’ was scored as ‘2’, ‘some of the time’ was scored as ‘3’, and ‘none of the time’ was scored as ‘4’). Responses to the first two questions (how often during the previous month they had felt calm or peaceful and happy) are reverse coded by the authors and added to the responses to the remaining three items (how often during the previous month they had felt nervous, down or blue, and depressed) to create a mental health scale of integer values ranging from five to twenty where a higher value indicates a better mental health status. The average mental health score was 15.5 with a standard deviation of 2.4 in our sample (Table 1, row 3). Individuals were also asked to evaluate their general health as excellent, very good, good, fair, and poor. A binary variable is created with a value of ‘1’ to indicate ‘good’ health or better and ‘0’ for ‘fair’ or ‘poor’ health. Approximately 93% of individuals reported good general health status (Table 1, row 4).
Three variables, one each for whether a respondent had ever consumed alcohol, ever smoked a cigarette, and had ever used marijuana, are created to indicate any substance use. For each of three indicator variables, a value of ‘1’ indicates that they answered in affirmative and ‘0’ if they answered in negative. Each of these three variables are examined separately. Approximately 73% individuals reported affirmatively to ever consuming alcohol, 42.7% reported affirmatively to ever smoking a cigarette, and 21.3% reported affirmatively to ever using marijuana (Table 1, rows 5–7). Similarly, indicator variables are created for whether a respondent had completed high school by the age of nineteen years and whether a respondent had enrolled in a two-year college or pursued higher degree at any point before the age of twenty-five years. In each case, a value of ‘1’ indicates that they answered in affirmative and ‘0’ if they answered in negative. Postsecondary education is only considered for those respondents who had completed high school since they were the only ones eligible for college enrollment. In our sample, 71.5% of participants had a high school diploma and 51% of participants had attended college (Table 1, rows 8–9). A respondent’s age when they entered the labor market full-time (more than twenty hours per week) for the first time while not simultaneously attending school is used as a continuous variable for measuring the timing of first employment. In our sample, the average age at first employment was around 21 years (Table 1, row 10).
Respondents were asked to report the number of times they were treated by a mental health professional during the past twelve months for emotional, mental, or psychiatric problems. Responses from the 2009 survey year were grouped into a dichotomous measure of mental health service use. It is assigned the value of ‘1’ if a respondent had reported seeking help at least once, and ‘0’ otherwise. Approximately 6% of respondents reported using of mental health service (Table 1, row 11). Respondents were asked to report the number of times they had an emotional, mental, or psychiatric problem that was not treated by a professional but resulted in their missing as least one full day of usual activities, such as work or school. A value of ‘1’ indicates that a respondent had experienced any life disrupting mental or emotional problem at least once that was not treated by a professional but resulted in their missing as least one full day of usual activities, and it is ‘0’ otherwise. Approximately 5% of respondents reported experiencing life-disrupting emotional or mental health problems (Table 1, last row).
Table 2 summarizes the outcome variables by victimization status. There are three categories of victimization status, which form the columns in Table 2 – never victimized, childhood victimization, and chronic victimization. In our sample, there was a monotonic decrease in mental health score (i.e., worsening of mental health) from among those who had never been bullied to those who experienced chronic victimization. Similarly, a larger percentage of respondents (94%) in the first category reported good general health status compared to 91% in the childhood victimization category and 87% in the chronic victimization category. Percentages of participants ever smoking a cigarette and ever using marijuana were highest among the third category (52% and 32%, respectively) and lowest among the first category (41% and 20%, respectively). While 79% of respondents in the ‘chronic victimization’ category reported ever consuming alcohol, the percent of ‘never victimized’ individuals who reported ever consuming alcohol (74%) was higher than the percent of ‘childhood victimized’ individuals who reported ever consuming alcohol (72%). With regard to educational attainment, 73% of individuals in the first category had high school diploma and 52% had attended college. Comparatively, a smaller percentage of individuals in the two victimization categories had these educational attainments – 65% and 45%, respectively, in the ‘childhood victimization’ category and 69% and 46%, respectively, in the ‘chronic victimization’ category. Age at first employment was similar across the three categories, approximately 20 years. Use of mental health services and experiencing life-disrupting emotional or mental problems were highest among those in the ‘chronic victimization’ category (12% and 11%, respectively), followed by those in the ‘childhood victimization’ category, and least among those who had never been bullied.
Table 2.
Descriptive statistics of outcome variables by bullying victimization category
| Dependent variables | Bullying victimization | ||
|---|---|---|---|
| Never | Childhood | Chronic | |
| Mental health score† |
15.70 (2.37) n = 4,152 |
15.35 (2.50) n = 832 |
14.53 (2.67) n = 247 |
| General health |
0.94 (0.24) n = 4,314 |
0.91 (0.29) n = 860 |
0.87 (0.34) n = 258 |
| Drinking |
0.74 (0.44) n = 4,271 |
0.72 (0.45) n = 852 |
0.79 (0.41) n = 258 |
| Smoking |
0.41 (0.49) n = 4,293 |
0.46 (0.50) n = 859 |
0.52 (0.50) n = 258 |
| Marijuana use |
0.20 (0.40) n = 4,247 |
0.24 (0.43) n = 847 |
0.32 (0.47) n = 256 |
| High school diploma |
0.73 (0.44) n = 4,858 |
0.65 (0.48) n = 960 |
0.69 (0.46) n = 278 |
| College attendance |
0.52 (0.50) n = 3,550 |
0.45 (0.50) n = 628 |
0.46 (0.50) n = 191 |
| Age at first employment, years |
20.90 (3.14) n = 4,667 |
20.61 (3.22) n = 928 |
20.47 (2.82) n = 268 |
| Mental health service use |
0.06 (0.23) n = 3,770 |
0.07 (0.25) n = 755 |
0.12 (0.33) n = 226 |
| Life-disrupting problems |
0.05 (0.21) n = 3,764 |
0.07 (0.26) n = 754 |
0.11 (0.32) n = 227 |
† The mental health score is measured on a Likert scale, ranging from 5 to 20 where a higher value indicates a better mental health status
Notes. Mental health score is a discrete variable and age at first employment is a continuous variable. Remaining variables are 0–1 indicator variables
Control Variables
Estimating equations, discussed in the next section, include multiple covariates or control variables consisting of various individual-level characteristics and household-level characteristics. Individual characteristics include gender, age at the time of the first survey, race-ethnicity (Black, Hispanic, and non-Black, non-Hispanic), body mass index (BMI) at the first survey, whether the respondent had any pre-existing learning difficulties (such as, dyslexia, attention disorder, or mental retardation), whether the respondent had any chronic health conditions (such as, asthma, diabetes, cancer, or anemia) and whether they had vision, hearing, or speech problems. Additionally, with regard to adverse experiences, respondents were asked if, between 2003 and 2008, they had experienced any violent crime (such as, physical or sexual assault, robbery, or arson), had been homeless for more than two nights in a row, whether their parents had divorced, whether a close relative had died, or if a household member had been unemployed for at least six months. A binary variable termed as ‘adverse experience’ is created and coded as ‘1’ if a respondent experienced at least one of the above-mentioned events, and it is ‘0’ if they had not experienced any of the above-mentioned adverse events. Whether a respondent had any kind of health insurance coverage in 2009 is indicated using a binary 0–1 variable. Household-level factors include number of children in the household, parental education, urban or rural residency, family structure, and family income. In a single-parent household, educational attainment of the residential parent is used. In a two-parent household, the highest level of educational attainment between the two parents is used. A respondent’s family structure at the initial survey could be one of four categories − two biological parents, one biological parent with one stepparent, single biological parent, and other family structure (adoptive parents, foster parents, or other relatives). Five binary variables are used to indicate whether the family-level income-to-poverty ratio is less than 100%, between 100% and 199%, between 200% and 399%, 400% or higher, or if income information is missing. The income-to-poverty ratio is a variable created by the NLSY97. It is a ratio comparing the gross family income to the federal poverty level for previous year, taking household size into account.
The control variables are summarized in Table 3. Data from 6,393 individuals, in the first round of the NLSY survey in 1997, are used. Survey attrition resulted in a sample size of 4,986 individuals by 2009. However, individual-level and household-hold characteristics of interest were similar in the 1997 and 2009 rounds. Approximately 25% of participants were Black, 19% were Hispanic, and 56% were of other race-ethnicity backgrounds. Approximately half of the sample were male. Average baseline BMI was around 22. Approximately 7% of participants reported a learning disability, 16% reported having trouble seeing, hearing, or speaking, and 10% reported having a chronic condition. In the 2009 survey, 61% of participants reported at least one adverse experience and 68% reported having health insurance coverage. Family income data were missing for around 16% of the households in each survey. Based on non-missing income data, a majority of the households were categorized in the 200-399% income-to-poverty ratio category. Parental educational attainment was around 13 years. Approximately half of the sample had both biological parents residing in the household and 32% were single-parent households. On average, sample households had between two and three children in residence. Approximately 71% of households were located in urban areas.
Table 3.
Descriptive statistics of covariates
| Variable | Frequency or Mean (SD) | |
|---|---|---|
| 1997 (n = 6,393) | 2009 (n = 4,986) | |
| Individual characteristics | ||
| Age in 1997 | 14.31 (1.46) | 14.26 (1.45) |
| Race | ||
| Black | 24.95% | 24.77% |
| Hispanic | 18.82% | 18.73% |
| Another race | 56.23% | 56.50% |
| Female | 48.01% | 50.00% |
| Male | 51.99% | 50.00% |
| Body Mass Index | 21.93 (4.56) | 21.96 (4.67) |
| Learning disability | 7.59% | 7.42% |
| Have trouble seeing, hearing, speaking | 16.36% | 16.85% |
| Have chronic conditions | 10.61% | 10.37% |
| Individual characteristics in adulthood | ||
| Adverse experience | 61.21% | |
| Health insurance coverage | 68.73% | |
| Household characteristics | ||
| Family income | ||
| Less than 100% of poverty | 17.57% | 17.35% |
| 100–199% of poverty | 17.41% | 17.87% |
| 200–399% of poverty | 29.81% | 29.96% |
| 400% plus of poverty | 18.41% | 18.73% |
| Missing poverty | 16.80% | 16.09% |
| Parental education, years | 13.32 (2.99) | 13.40 (3.00) |
| Family structure | ||
| Two biological parents | 51.49% | 52.33% |
| One biological parent and one stepparent | 14.08% | 13.54% |
| Single parent | 32.69% | 32.59% |
| Other family structure | 1.74% | 1.54% |
| Number of household members under the age of 18 | 2.42 (1.24) | 2.44 (1.24) |
| Urban | 71.41% | 71.08% |
| Rural | 28.59% | 28.92% |
Methodology
The effects of childhood and chronic victimization on young adults’ outcomes are estimated using propensity score matching (PSM) regression models. Ordinary least squares (OLS) models are also estimated for comparison purposes. The PSM point estimate is the difference in mean outcomes between the treatment and control groups, while the OLS point estimates come from modeling the outcome. In other words, in the PSM method, victimized individuals are matched with never victimized individuals with similar measured covariates, while unmatched observations are discarded. The association between victimization and the outcome variable are assessed using the matched pairs with similar characteristics, and any difference in the outcome variables are attributed to victimization. In theory, matching produces less biased estimates of the association between an exposure or treatment (bullying victimization, in this study) and outcomes than does unmatched OLS regression, particularly when the overlap in characteristics between victimized and non-victimized groups is poor (Vable et al., 2019). However, neither the OLS results nor the PSM results should be interpreted as causal relationships between victimization and the outcome variables. PSM, simply, reduces many covariates into a single score and can increase efficiency over OLS regressions (Rubin & Thomas, 2000).
While randomized controlled trials (RCTs) are considered the gold standard method for estimating effects of treatments or interventions, experimental data collection is either not possible or infeasible in many research studies. PSM is a widely used technique to estimate treatment effects in observational studies (Abadie and Imben, 2016; Imbens 2015). In this study, the treatment variables are childhood and chronic bullying victimization, and understanding the impact of bullying victimization on the outcome variables is of interest. A treatment is assigned by randomization in RCTs, and the average treatment effect (ATE) is directly computed from the experimental data. In observational studies, however, the treated individuals often systematically differ from the untreated individuals. Thus, an unbiased estimate of the average treatment effect cannot be directly computed; rather, the propensity score can be used to estimate the average treatment effect (Rosenbaum and Rubin, 1983). The propensity score is the probability of treatment assignment conditional on observed baseline characteristics or the control variables (Austin, 2011). In other words, the likelihoods of childhood and chronic bullying victimization are, first, estimated as functions of the control variables (also known as baseline covariates). The control variables are then, also, included in the final regression models.
Conditional on the propensity score, victimized (or treated) and never victimized (or untreated) individuals are matched. ATT is the average effect of treatment (in this case, bullying victimization) on the individuals who experienced bullying. In this study, the average treatment effect on the treated (ATT) is of greater interest than the ATE. Thus, once matched treated and control pairs are formed, the treatment or bullying victimization effect is obtained by calculating the differences in outcomes between the treated individuals and their matches.
Unbiased estimates of average treatment effects are achieved under two conditions (Rosenmbaum and Rubin, 1983). First, treatment assignments must be independent of the potential outcome(s) conditional on the observed baseline covariates. Second, every individual must have a nonzero probability of being assigned in either treatment or control group. The propensity score analysis is performed using ‘MatchIt’ in R and the nearest neighbor algorithm with replacement and caliper distance equal to 0.2 of the standard deviation of the logit of the propensity score is applied. Appendix 1 compares the standardized mean differences in the baseline covariates in the unmatched and matched samples. Smaller differences (in absolute terms) are observed in the matched sample compared to the unmatched sample indicating balance in measured variables between victimized and non-victimized individuals in the matched sample. In the propensity score model, a linear relationship between the control variables and the log-odds of bullying victimization is assumed. Proportion of treatment and control individuals across distributions of propensity scores are shown in Appendix 2. For comparison purposes, both control and treatment propensity scores are shown in the same figure and the proportions of treatment group individuals are shown on an inverted scale. The figures show similar proportion of treatment and control individuals across the linearized propensity scores distributions. Balance diagnostics shown in Appendix 1 and Appendix 2 assess whether the distribution of baseline covariates is similar between victimized and non-victimized groups. When point estimates from OLS regression and matching methods are similar (i.e., confidence intervals overlap), there is no efficiency gain in estimating PSM models. On the other hand, if the PSM and OLS estimates are dissimilar then matching inferences are unbiased compared with OLS inferences (Vable et al., 2019).
Results
Table 4 shows the estimated associations between bullying victimization and the health and educational attainment outcomes. Following the balance diagnostics (Appendices 1 and 2), we interpret estimates from the PSM models only. However, estimated coefficients from the OLS models are also shown in Table 4. There are two treatment groups − victims of childhood bullying and victims of chronic bullying. The control group consists of individuals who were never victimized in childhood or adolescence. In the PSM model, bullying victimization is negatively associated with self-reported mental health score and general health status. The mental health measure is a discrete variable obtained from adding five distinct survey items each on a four-point Likert scale. The mental health scale ranges from five to twenty (integer values), with a lower value reflecting worse mental health. Among both childhood and chronic victims, the relationship between bullying victimization and mental health is negative, reflecting worse mental health among victimized individuals compared to non-victimized individuals, on average. The average mental health score is a half-point higher among those who had experienced childhood bullying victimization, while it is a full point higher among those who had experienced chronic bullying victimization compared to non-victimized individuals. The differences are statistically significant at 99% confidence level. Similarly, with regard to general health status, the likelihood of reporting good general health is 4.2% point lower among those who had experienced childhood bullying victimization and 7.4% point lower among those who had experienced chronic bullying victimization compared to non-victimized individuals. These two estimated differences between the non-victimized and victimized groups are statistically significant at 99% confidence level.
Table 4.
The effects of bullying victimization on health outcomes and educational attainment
| Dependent variable | Childhood bullying victimization | Chronic bullying victimization | ||
|---|---|---|---|---|
| OLS | PSM | OLS | PSM | |
| Mental health score | ‒0.392*** (0.091) | ‒0.502*** (0.127) | ‒1.166*** (0.156) | ‒1.182*** (0.232) |
| General health | ‒0.029*** (0.009) | ‒0.042*** (0.013) | ‒0.071*** (0.015) | ‒0.074*** (0.026) |
| Drinking | ‒0.010 (0.016) | ‒0.002 (0.024) | 0.022 (0.027) | 0.066* (0.040) |
| Smoking | 0.038** (0.018) | 0.047* (0.025) | 0.066** (0.031) | 0.105** (0.043) |
| Marijuana use | 0.041*** (0.015) | 0.039* (0.022) | 0.103*** (0.026) | 0.078* (0.042) |
| High school diploma | ‒0.036** (0.015) | ‒0.013 (0.023) | ‒0.022 (0.025) | ‒0.047 (0.04) |
| College attendance | 0.005 (0.018) | ‒0.005 (0.026) | ‒0.004 (0.030) | ‒0.026 (0.042) |
| Age at first employment, years | ‒0.112 (0.108) | ‒0.332** (0.166) | ‒0.380** (0.187) | ‒0.522* (0.277) |
*** p < 0.01; ** p < 0.05; * p < 0.1
Standard errors are shown in parenthesis
Notes. Mental health score is a discrete variable, and a higher score indicates better mental health. A negative coefficient for the ‘mental health score’ outcome implies worse mental health associated with bullying victimization compared to no victimization. Age at first employment is a continuous variable. A negative coefficient for the ‘age at first employment’ outcome implies younger age at labor market entry among those who experienced bullying victimization compared to those never victimized. Remaining variables are 0–1 indicator variables. A negative (positive) coefficient for such an outcome implies lower (greater) likelihood of that outcome among those who experienced bullying victimization compared to those never victimized
In examining risky health behaviors, the likelihood of ever consuming alcohol is not statistically significantly different between the non-victimized and childhood victimization groups. The likelihoods of ever smoking a cigarette and marijuana use are higher by 4.7% point and 3.9% point, respectively, in the childhood victimization group. Both estimates are statistically significant at 90% confidence level in the PSM models. Estimated associations between chronic victimization and likelihoods of the three risky health behaviors are larger. The likelihoods of ever consuming alcohol, smoking a cigarette, and using marijuana are higher by 6.6% point (statistically significant at 90% confidence level), 10.5% point (statistically significant at 95% confidence level), and 7.8% point (statistically significant at 90% confidence level), respectively, in the chronic victimization group compared to the never victimized group.
Educational attainment, in terms of receiving high school diploma and attending college, while negatively associated with victimization, is not statistically significantly different between the victimized and non-victimized groups in the PSM models. Both childhood and chronic victimization are associated with earlier labor market entry. Victims of chronic bullying tend to join the labor market at a younger age, approximately six months sooner in the PSM model, and the estimated coefficient is statistically significant at 90% confidence level. Victims of childhood bullying join approximately four months sooner compared to non-victimized individuals, and the estimated coefficient is statistically significant at 95% confidence level.
In examining use of mental health services and incidences of emotional or mental health problems in adulthood, additional control variables are used as described in the ‘Data’ section. They include any adverse experience in adulthood (violent crime, homelessness, parental divorce, death of a close relative, unemployment of family member), health insurance coverage, and mental health measure from the 2004 survey. The results are presented in Table 5. In the PSM models, childhood bullying victimization is associated with a 2.8% point greater likelihood of experiencing life-disrupting emotional or mental health problems compared to non-victims, and the estimated coefficient is statistically significant at 95% confidence level. Point estimates, as with other outcome variables, are larger in the case of chronic bullying victimization. Chronic bullying victimization is associated with an 8.8% point greater likelihood of experiencing life-disrupting emotional or mental health problems compared to non-victims, which is statistically significant at 99% confidence level, and a 5.8% point higher likelihood of utilizing mental health services, which is statistically significant at 95% confidence level.
Table 5.
The effects of bullying victimization on the use of mental health services and emotional problems
| Childhood bullying victimization | Chronic bullying victimization | |||
|---|---|---|---|---|
| OLS | PSM | OLS | PSM | |
| Life-disrupting emotional/mental problems | 0.022** (0.009) | 0.028** (0.013) | 0.053*** (0.015) | 0.088*** (0.024) |
| Use of mental health services | 0.008 (0.009) | 0.004 (0.014) | 0.043** (0.017) | 0.058** (0.029) |
*** p < 0.01; ** p < 0.05; * p < 0.1
Standard errors are shown in parenthesis
Notes. Both variables are 0–1 indicator variables. A positive coefficient for such an outcome implies greater likelihood of that outcome among those who experienced bullying victimization compared to those never victimized
The difference in outcomes between chronic victims and non-victims and the difference in outcomes between childhood victims and non-victims are compared, and the PSM differentials are shown in Table 6. In other words, a differential is estimated as the estimated difference between chronic victims and non-victims for a specific outcome minus the estimated difference between childhood victims and non-victims for that outcome. A negative differential indicates a larger estimated association between victimization and an outcome among chronic victims compared to childhood victims when both estimates are negative and the point estimate (in absolute value) among chronic victims is larger. A positive differential indicates a larger estimated association among chronic victims compared to childhood victims when both estimates are positive and the point estimate (in absolute value) among chronic victims is larger. The estimated differentials are statistically significant for three of the ten outcome variables – mental health score, general health status, and use of mental health services.
Table 6.
Estimated differential effects of bullying victimization
| Dependent variable | Estimated differential |
|---|---|
| Mental health score | ‒0.757*** (0.247) |
| General health | ‒0.062** (0.030) |
| Drinking | 0.066 (0.042) |
| Smoking | 0.035 (0.049) |
| Marijuana use | 0.043 (0.048) |
| High school diploma | 0.014 (0.045) |
| College attendance | 0.021 (0.051) |
| Age at the first employment, years | ‒0.216 (0.333) |
| Life-disrupting emotional/mental problems | 0.018 (0.038) |
| Use of mental health services | 0.067** (0.029) |
*** p < 0.01; ** p < 0.05; * p < 0.1
Standard errors are shown in parenthesis
Notes. Estimated differential is the difference in two differences – the difference in outcome between longer-term victims and non-victims and the difference in outcome between shorter-term victims and non-victims using PSM models. A negative differential for the ‘mental health score’ outcome implies worse mental health associated with chronic bullying victimization compared to non-victimization than childhood bullying victimization compared to non-victimization. A positive (negative) differential for an indicator variable outcome implies higher (lower) likelihood of the outcome in the chronic bullying victimization group compared to non-victimization than in the childhood bullying victimization group compared to non-victimization
In the case of mental health score, the differential is negative and statistically significant at 99% confidence level. Recall that a higher mental health score indicates better mental health status and that previously (Table 4) it was found that bullying victimization is associated with lower mental health score (i.e., poorer mental health). The negative differential indicates that the magnitude (in absolute value) of the lower mental health score among those in the childhood victimization group compared to the non-victimized group was smaller than the magnitude of the lower mental health score among those in the chronic victimization group compared to the non-victimized group. In other words, relative mental health score was worse among those in the chronic victimization group compared to those in the childhood victimization group, and the differential was between a half and a full point on the Likert scale.
General health status is an indicator variable. Previously, we had found that the likelihood of reporting good general health was lower (or negative) in each of the victimization group compared to never victimized individuals (Table 4). The negative and statistically significant (at 95% confidence level) differential in Table 6 implies that the likelihood of reporting good general health is relatively lower in the chronic victimization group, by 6.2% point, compared to the childhood victimization group. In other words, while both victimization groups reported lower likelihood of good general health compared to never victimized individuals, the likelihood of good general health is lower among chronic victims compared to childhood victims.
Lastly, use of mental health services is greater among chronic victims than among childhood victims compared to those who had never been victimized. Previously, we had found that the likelihood of use of mental health services was statistically and significantly higher among the chronic bullying victimization group compared to never victimized individuals, and that there was not a statistical difference among the childhood bullying victimization group compared to never victimized individuals. The positive differential implies that the greater use of mental health services among the chronic bullying victimization group (compared to non-victims) was 6.7% point higher compared to the difference in use of mental health services between the childhood bullying victimization group and non-victims. The estimated differential is statistically significant at 95% confidence level. Remaining differentials are not different at a statistically significant level. In other words, while chronic bullying victimization appears to have a stronger association (in terms of absolute value of point estimates) with all outcome variables examined in this study, only in three outcome variables are the mean differences in outcomes larger (in absolute value) and statistically significant among chronic victims than among childhood victims compared to non-victims. Thus, there is a statistically significant difference between chronic versus childhood bullying victimization with regard to mental health, general health status, and use of mental health service.
Discussion and Conclusions
Bullying is an important social and public health issue. Previous research has mainly focused on short-term bullying, mostly in childhood, and the associated consequences. This study attempts to address the question of how and to what extent different periods of exposure to bullying victimization affect mental and physical health status, certain risky health behaviors, economic well-being, and utilization of mental health services in adulthood. In other words, this study compares and contrasts how childhood and chronic bullying victimization pre-adulthood are associated with various outcomes in early adulthood. One limitation of this study in the definition of childhood and chronic bullying victimization. Bullying is considered chronic when a person experienced bullying both before the age of twelve years and between twelve and eighteen years while childhood bullying refers to a child who experienced bullying only before the age of twelve years and not in later years. These bullying variables are dichotomous, so it is difficult to measure the intensity or frequency of bullying. Even though the questions asked about repeated bullying, the exact frequency of victimization is unknown. Ideally, we would have access to how often the person experienced bullying and to what extent to define a more comprehensive measure of victimization. The definition of chronic bullying victimization could be improved if there was information on bullying victimization for each year and the frequency of victimization per year. In this study, the measure of chronic bullying victimization provides a limited dose-response view of respondents being bullied throughout two developmental periods. Having a more comprehensive measure of victimization would allow us to better assess the dose effect. Our findings seem to indicate that that more frequent and prolonged victimization is associated with higher risk of mental health related difficulties in adulthood.
Study results suggest that it is important not only to analyze childhood bullying victimization but also chronic bullying victimization, since the adverse effects of chronic bullying victimization are frequently larger. In this study, childhood bullying victimization is associated with poorer mental and general health, greater propensity to smoke and use marijuana, earlier entry into the labor market, and greater likelihood to experience emotional problems. In addition to these associations, chronic bullying victimization is related to greater likelihoods of drinking and utilization of mental health services in adulthood. OLS and PSM models are estimated. There is no efficiency gain with matching methods when the point estimates from the two types of regression models are similar. However, in most cases, the point estimates are found to be dissimilar, in which case matching inferences are unbiased as compared to OLS inferences.
A shortcoming of the methodology is the inability to control for any unobserved characteristic that may simultaneously influence the treatment exposure and potential outcomes. Such unobserved characteristics may bias estimators, either under- or overestimating the effects of treatment exposure. Misreporting or measurement error may arise due to “victim mentality” (overreporting) or shame and embarrassment (underreporting). Misclassification due to underreporting leads to downward bias of the effects of bullying. Endogeneity issues are also of concern. For instance, body mass index (BMI) is one of the control variables, which could lead to selection issues in that bullying could lead to body weight changes (Sutin et al., 2016) and that bullying could be more prevalent among overweight and obese children (Lumeng et al., 2010). Addressing endogeneity concerns is outside the scope of this paper; thus, only the baseline BMI is used as a control variable instead of change in BMI between the two data collection points.
While the durations of childhood and chronic bullying victimization, as used in this study, are based on available survey data and not clinical definitions, the findings from our study show that chronic bullying victimization induces larger negative effects. To mitigate adverse effects of chronic bullying, prevention efforts need to start early. Investing in anti-bullying programs and improvements in access to health services could alleviate the long-term impacts on mental health and productivity associated with bullying victimization. Robust estimates of impact of bullying may be estimated with richer long-run panel dataset that capture bullying related measures at every survey. While the authors are not aware of any database that collects information regarding history of bullying victimization, such information can improve our understanding of the impact of victimization on health and productivity in childhood, adolescence, and adulthood, and of developing effective interventions for different age groups to alleviate the trauma of bullying victimization.
Electronic Supplementary Material
Below is the link to the electronic supplementary material.
Human Participant Protection
Because this study used only publicly available data, it was not subject to regulation by the institutional review board in any of author’s university.
Conflict of interest
Authors do not have any conflict of interest.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- Abadie A, Imbens GW. Matching on the estimated propensity score. Econometrica. 2016;84(2):781–807. doi: 10.3982/ECTA11293. [DOI] [Google Scholar]
- Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research. 2011;46(3):399–424. doi: 10.1080/00273171.2011.568786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Baiden P, Tadeo SK. Investigating the association between bullying victimization and suicidal ideation among adolescents: Evidence from the 2017 Youth Risk Behavior Survey. Child Abuse and Neglect. 2020;102:104417. doi: 10.1016/j.chiabu.2020.104417. [DOI] [PubMed] [Google Scholar]
- Bouffard LA, Koeppel MD. Understanding the potential long-term physical and mental health consequences of early experiences of victimization. Justice Quarterly. 2014;31(3):568–587. doi: 10.1080/07418825.2012.734843. [DOI] [Google Scholar]
- Boynton-Jarrett R, Hair E, Zuckerman B. Turbulent times: Effects of turbulence and violence exposure in adolescence on high school completion, health risk behavior, and mental health in young adulthood. Social Science and Medicine. 2013;95:77–86. doi: 10.1016/j.socscimed.2012.09.007. [DOI] [PubMed] [Google Scholar]
- Brendgen M, Poulin F. Continued bullying victimization from childhood to young adulthood: A longitudinal study of mediating and protective factors. Journal of Abnormal Child Psychology. 2018;46(1):27–39. doi: 10.1007/s10802-017-0314-5. [DOI] [PubMed] [Google Scholar]
- Brimblecombe N, Evans-Lacko S, Knapp M, King D, Takizawa R, Maughan B, Arseneault L. Long term economic impact associated with childhood bullying victimisation. Social Science and Medicine. 2018;208:134–141. doi: 10.1016/j.socscimed.2018.05.014. [DOI] [PubMed] [Google Scholar]
- Brown S, Taylor K. Bullying, education and earnings: Evidence from the National Child Development Study. Economics of Education Review. 2008;27(4):387–401. doi: 10.1016/j.econedurev.2007.03.003. [DOI] [Google Scholar]
- Brunstein Klomek A, Barzilay S, Apter A, Carli V, Hoven CW, Sarchiapone M, Hadlaczky G, Balazs J, Kereszteny A, Brunner R, Kaess M, Bobes J, Saiz PA, Cosman D, Haring C, Banzer R, McMahon E, Keeley H, Kahn JP, Potuvan V, Podlogar T, Sisask M, Varnik A, Wasserman D. Bi-directional longitudinal associations between different types of bullying victimization, suicide ideation/attempts, and depression among a large sample of european adolescents. Journal of Child Psychology and Psychiatry. 2019;60(2):209–215. doi: 10.1111/jcpp.12951. [DOI] [PubMed] [Google Scholar]
- Connolly EJ, Kavish N, Cooke EM. Testing the causal hypothesis that repeated bullying victimization leads to lower levels of educational attainment: A sibling-comparison analysis. Journal of School Violence. 2019;18(2):272–284. doi: 10.1080/15388220.2018.1477603. [DOI] [Google Scholar]
- Eriksen TLM, Nielsen HS, Simonsen M. Bullying in elementary school. Journal of Human Resources. 2014;49(4):839–871. doi: 10.1353/jhr.2014.0039. [DOI] [Google Scholar]
- Evans CB, Smokowski PR, Rose RA, Mercado MC, Marshall KJ. Cumulative bullying experiences, adolescent behavioral and mental health, and academic achievement: An integrative model of perpetration, victimization, and bystander behavior. Journal of Child and Family Studies. 2019;28(9):2415–2428. doi: 10.1007/s10826-018-1078-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Evans-Lacko S, Takizawa R, Brimblecombe N, King D, Knapp M, Maughan B, Arseneault L. Childhood bullying victimization is associated with use of mental health services over five decades: A longitudinal nationally representative cohort study. Psychological Medicine. 2017;47(1):127–135. doi: 10.1017/S0033291716001719. [DOI] [PubMed] [Google Scholar]
- Fernandez, C. A., Christ, S. L., LeBlanc, W. G., Arheart, K. L., Dietz, N. A., McCollister, K. E., Muntaner, C., Muennig, P., & Lee, D. J. (2015). Effect of childhood victimization on occupational prestige and income trajectories.PloS one, 10(2), e0115519. [DOI] [PMC free article] [PubMed]
- Geoffroy MC, Boivin M, Arseneault L, Turecki G, Vitaro F, Brendgen M, Renaud J, Séguin JR, Tremblay RE, Côté SM. Associations between peer victimization and suicidal ideation and suicide attempt during adolescence: Results from a prospective population-based birth cohort. Journal of the American Academy of Child and Adolescent Psychiatry. 2016;55(2):99–105. doi: 10.1016/j.jaac.2015.11.010. [DOI] [PubMed] [Google Scholar]
- Hoffman CY, Phillips MD, Daigle LE, Turner MG. Adult consequences of bully victimization: Are children or adolescents more vulnerable to the victimization experience? Youth Violence and Juvenile Justice. 2017;15(4):441–464. doi: 10.1177/1541204016650004. [DOI] [Google Scholar]
- Imbens GW. Matching methods in practice: Three examples. Journal of Human Resources. 2015;50(2):373–419. doi: 10.3368/jhr.50.2.373. [DOI] [Google Scholar]
- Lebrun-Harris LA, Sherman LJ, Miller B. State-level prevalence of bullying victimization among children and adolescents, National Survey of Children’s health, 2016–2017. Public Health Reports. 2020;135(3):303–309. doi: 10.1177/0033354920912713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee KS, Vaillancourt T. Longitudinal associations among bullying by peers, disordered eating behavior, and symptoms of depression during adolescence. JAMA Psychiatry. 2018;75(6):605–612. doi: 10.1001/jamapsychiatry.2018.0284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lereya ST, Copeland WE, Costello EJ, Wolke D. Adult mental health consequences of peer bullying and maltreatment in childhood: Two cohorts in two countries. The Lancet Psychiatry. 2015;2(6):524–531. doi: 10.1016/S2215-0366(15)00165-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lumeng JC, Forrest P, Appugliese DP, Kaciroti N, Corwyn RF, Bradley RH. Weight status as a predictor of being bullied in third through sixth grades. Pediatrics. 2010;125:e1301–e1307. doi: 10.1542/peds.2009-0774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mojaverian T, Hashimoto T, Kim HS. Cultural differences in professional help seeking: A comparison of Japan and the US. Frontiers in Psychology. 2013;3:615. doi: 10.3389/fpsyg.2012.00615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mojtabai R. Trends in contacts with mental health professionals and cost barriers to mental health care among adults with significant psychological distress in the United States: 1997–2002. American Journal of Public Health. 2005;95(11):2009–2014. doi: 10.2105/AJPH.2003.037630. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Oliveira FR, de Menezes TA, Irffi G, Oliveira GR. Bullying effect on student’s performance. EconomiA. 2018;19(1):57–73. doi: 10.1016/j.econ.2017.10.001. [DOI] [Google Scholar]
- Olweus D. Bullying at school: What we know and what we can do. Malden, MA: Blackwell Publishing; 1993. [Google Scholar]
- Ponzo M. Does bullying reduce educational achievement? An evaluation using matching estimators. Journal of Policy Modeling. 2013;35(6):1057–1078. doi: 10.1016/j.jpolmod.2013.06.002. [DOI] [Google Scholar]
- Quinn ST, Stewart MC. Examining the long-term consequences of bullying on adult substance use. American Journal of Criminal Justice. 2018;43(1):85–101. doi: 10.1007/s12103-017-9407-5. [DOI] [Google Scholar]
- Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. doi: 10.1093/biomet/70.1.41. [DOI] [Google Scholar]
- Rubin DB, Thomas N. Combining propensity score matching with additional adjustments for prognostic covariates. Journal of the American Statistical Association. 2000;95(450):573–585. doi: 10.1080/01621459.2000.10474233. [DOI] [Google Scholar]
- Sarzosa M, Urzúa S. Bullying among adolescents: The role of skills. Quantitative Economics. 2021;12(3):945–980. doi: 10.3982/QE1215. [DOI] [Google Scholar]
- Schnyder N, Panczak R, Groth N, Schultze-Lutter F. Association between mental health-related stigma and active help-seeking: Systematic review and meta-analysis. The British Journal of Psychiatry. 2017;210(4):261–268. doi: 10.1192/bjp.bp.116.189464. [DOI] [PubMed] [Google Scholar]
- Sigurdson JF, Undheim AM, Wallander JL, Lydersen S, Sund AM. The long-term effects of being bullied or a bully in adolescence on externalizing and internalizing mental health problems in adulthood. Child and Adolescent Psychiatry and Mental Health. 2015;9(1):1–13. doi: 10.1186/s13034-015-0075-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Singham T, Viding E, Schoeler T, Arseneault L, Ronald A, Cecil CM, McCrory E, Rijsdijk F, Pingault JB. Concurrent and longitudinal contribution of exposure to bullying in childhood to mental health: The role of vulnerability and resilience. JAMA Psychiatry. 2017;74(11):1112–1119. doi: 10.1001/jamapsychiatry.2017.2678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stewart-Tufescu A, Salmon S, Taillieu T, Fortier J, Afifi TO. Victimization experiences and mental health outcomes among grades 7 to 12 students in Manitoba, Canada. International Journal of Bullying Prevention. 2021;3(1):1–12. doi: 10.1007/s42380-019-00056-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sutin AR, Robinson E, Daly M, Terracciano A. Parent-reported bullying and child weight gain between ages 6 and 15. Childhood Obesity. 2016;12(6):482–487. doi: 10.1089/chi.2016.0185. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Takizawa R, Maughan B, Arseneault L. Adult health outcomes of childhood bullying victimization: Evidence from a five-decade longitudinal british birth cohort. American Journal of Psychiatry. 2014;171(7):777–784. doi: 10.1176/appi.ajp.2014.13101401. [DOI] [PubMed] [Google Scholar]
- Vable AM, Glymour KMV, Rigdon MM, Drabo J, Basu S. Performance of matching methods as compared with unmatched ordinary least squares regression under constant effects. American Journal of Epidemiology. 2019;188(7):1345–1354. doi: 10.1093/aje/kwz093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- National Center for Education Statistics (2020). Bureau of Justice Statistics. Indicators of school crime and safety: 2019. Washington, DC: 2020
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.