Abstract
Adolescent employment during high school has become the norm in the United States, but studies of associated outcomes have yielded mixed results. These discrepant findings may be partly attributable to study methods, including differences in how adolescent employment is measured and how selection factors are taken into account. The present study, based on data from the Child Development Supplement of the Panel Study of Income Dynamics, aims to continue untangling these complexities by (a) examining whether the strength of theoretical predictors varies when predictors are assessed in a comprehensive model that simultaneously controls several psychological, family, and community factors; (b) determining whether the strength of predictors varies depending on how adolescent employment is measured (work status, work duration, and work intensity); and (c) assessing whether race moderates some of these relationships. Results indicate differences in how each predictor is related to each dimension of adolescent employment, as well as a moderating effect of race on the relationship between educational expectations and number of hours adolescents worked each week.
Being employed during high school has become the norm in the United States, as a majority of adolescents report working for pay during the school year at some point prior to leaving high school (Staff, Messersmith, & Schulenberg, 2009). Studies of youth outcomes associated with adolescent employment have yielded a mixed picture. Some have found that adolescent employment is associated with indicators of positive development such as greater employability, increased future earnings, and higher occupational status (e.g., Ruhm, 1997), whereas others have linked adolescent employment to negative outcomes such as antisocial behavior (e.g., Wright & Cullen, 2004), lower academic performance (e.g. Mortimer, Finch, Ryu, Shanahan, & Call, 1996; Steinberg & Dornbusch, 1991), and lower educational attainment and occupational status (e.g., Marsh, 1991; Nagengast, Marsh, Chiorri, & Hau, 2014). This pattern of inconsistency has prompted study of factors that moderate links between adolescent work and youth outcomes.
The most frequently examined and robust moderator found in this research is number of hours worked, with low-intensity work (<20 hours/week) typically associated with positive youth outcomes and high-intensity work (>20 hours/week) associated with negative outcomes (Staff et al., 2009). Sociodemographic factors such as parent education level and race/ethnicity have also been found to influence the relation between adolescent employment and youth outcomes, although the evidence is not altogether consistent. In Staff and Mortimer’s (2008) longitudinal study of youth in Minnesota, steady work (high duration, low intensity) during high school was more positively associated with postsecondary educational attainment among youth whose parents had low levels of education (as compared with youth whose parents had high levels of education), although this moderation effect did not hold in subsequent research when prior educational performance (e.g., grade point average [GPA] and math achievement) was controlled (Hwang & Domina, 2017). Other evidence suggests that any costs possibly attributable to long hours of student work are most severe for those who are most advantaged (Bachman, Staff, O’Malley, & Freedman-Doan, 2013). Whereas high-intensity work during high school was found to be negatively associated with high school GPA, college enrollment, attainment of college certificate, and attainment of an associate degree among White students, it was unrelated to these outcomes among Black students (Bachman et al., 2013; Hwang & Domina, 2017). In addition, research indicates that students with the most highly educated parents have the strongest negative relations between work intensity and GPA (Bachman et al., 2013).
Evidence that race and socioeconomic status (SES) influence the relation between adolescent employment and youth outcomes has raised questions about differential selection into employment, that is, whether employed Black and low-SES adolescents have different characteristics than their nonemployed counterparts even before they begin to work (e.g., motivation, social skills), and whether such characteristics are stronger determinants of employment and access to hour-intensive jobs for Black and lower SES youth than White and more affluent youth. Differential selection into employment, perhaps driven in part by greater obstacles that Black and lower SES adolescents face in obtaining employment, compared with White and more affluent adolescents, may result in differential quality of employment and differential outcomes of employment by race and social class (Bachman et al., 2013; Hwang & Domina, 2017).
Researchers have long questioned the extent to which associations between adolescent employment and youth outcomes reflect causation or selection effects or some of each (Bachman et al., 2013; Staff et al., 2009). Differences in whether and how studies address preexisting characteristics that influence adolescents’ propensity to work appear to be a major contributor to the pattern of inconsistency in findings assessing links between adolescent employment and youth outcomes (Monahan, Lee, & Steinberg, 2011). Recent rigorous research demonstrating the robust moderating influence of race on the link between adolescent employment and youth outcomes further underscores why addressing differential selection into employment is essential to advance our understanding of the effects of adolescent employment (Bachman et al., 2013; Hwang & Domina, 2017).
The present study had two major goals: (a) to examine links between three dimensions of adolescent employment (work status, work duration, and work intensity) and several psychological, family, and community factors and (b) to determine whether some of these links are conditional on race. When possible and relevant from a theoretical standpoint, we assess psychological/family variables measured during childhood and early adolescence in Wave 1 (ages 6-13 years) when child labor laws prohibit participation in formal paid employment and prior to the time that youth typically begin working for pay. Comparatively few studies have examined characteristics and processes during this developmental period as predictors of adolescent employment. Similar to prior work, we focus on employment during the school year, rather than year-round employment. The predictors and correlates of school-year employment versus summer employment likely differ, given the challenges of working during the school year. In addition, the importance of individual psychosocial factors as predictors of adolescent work may be magnified during the school year, compared with the summer months, when adolescents generally are less likely to be working (Marsh, 1991; Staff & Mortimer, 2008).
Precursors of Adolescent Work
Drawing on Bronfenbrenner’s (1989) ecological systems framework, we examined three levels of variables as precursors of adolescent work—psychological factors (i.e., self-reliance, behavioral problems, and expectations of 4-year college completion), family context (i.e., domestic chores, parent educational expectations, middle-class cultural socialization), and community context (i.e., adolescent social capital, neighborhood social organization). Research relevant to each of these categories of variables is discussed below.
Psychological Factors
Quantitative research employing prospective longitudinal designs has identified a few individual-level characteristics associated with an increase in the odds of adolescent work or certain patterns of adolescent work. In samples of low-income youth, greater autonomy (parent-reported) during middle childhood/mid-adolescence predicted steady adolescent employment 3 years later (Purtell & McLoyd, 2011), while grade retention during middle childhood (maternal report) and coming from families that were on welfare longer predicted lower odds of adolescent employment (Leventhal, Graber, & Brooks-Gunn, 2001). Greater odds of unemployment during the transition to adulthood has been linked to lower reading scores and delinquency during adolescence (Caspi, Wright, Moffitt, & Silva, 1998), and greater odds of unemployment from adolescence to midlife has been associated with low self-control during mid-adolescence (Daly, Delaney, Egan, & Baumeister, 2015).
Qualitative research also implicates behavior and academic competence as factors that increase the odds of employment among low-income youth. Newman (1999) found that the ability to communicate and work well with others were key criteria employers used to select candidates from applicant pools of youth residing in Harlem. Employers in Wilson’s (1996) Chicago-based study reported that basic academic skills and “soft skills” such as dependability, a strong sense of responsibility, and a strong desire to work hard were pivotal factors in their decisions about hiring low-income youth. They valued middle-class cultural capital—a resource that encompasses ways of speaking and dressing (Dumais, 2002)—noting that use of “street talk” instead of proper grammar, “inappropriate” dress, and other forms of self-presentation hurt low-income males in job interviews. Low levels of middle-class cultural capital is also a theme in MacLeod’s (2009) seminal ethnographic work detailing the employment struggles of low-income Black and White male youth.
There is consistent evidence linking lower academic aspirations to an increased propensity to work intensively (e.g., Entwisle, Alexander, & Olson, 2005; Lee & Staff, 2007; Steinberg, Fegley, & Dornbusch, 1993). Lower investment in education among youth who take on intensive paid work provides a highly plausible explanation for this finding (e.g., Bachman et al., 2013; Steinberg & Dornbusch, 1991). Although academic aspirations and intensive work are inversely related generally, intensive adolescent work positively predicts educational aspirations and actual college attendance when students report that their reason for working is to save money for college (Marsh, 1991; Marsh & Kleitman, 2005; Nagengast et al., 2014). In the present study, we examined postsecondary educational expectations (i.e., beliefs about future outcomes), rather than educational aspirations (i.e., what an individual ideally would like to achieve in the future), because the former, compared with the latter, is a stronger predictor of behavioral choices associated with academic success and attainment (Wood, Kurtz-Costes, & Copping, 2011). This suggests, for example, that educational expectations might be a stronger predictor of employment intensity than educational aspirations.
The present study tested several hypotheses about the contribution of psychological factors to adolescent employment that are informed by prior research. We hypothesized that higher levels of self-reliance during childhood and early adolescence would be positively associated with employment status and employment duration 5 years later (Purtell & McLoyd, 2011). We also posited that higher levels of behavioral problems during childhood and early adolescence would be linked longitudinally to lower odds of employment and lower duration and intensity of adolescent employment (Caspi et al., 1998; Daly et al., 2015). In light of the positive association between steady work (high duration, low intensity) and postsecondary educational attainment found in prior research (Staff & Mortimer, 2008), we expected that higher postsecondary educational expectations would be concurrently associated with higher odds of employment and longer duration of employment, but lower intensity of employment.
Family Context
There is a surprising paucity of research investigating aspects of family context as precursors of adolescent employment. Spending more time doing household chores during childhood (Entwisle, Alexander, Olson, & Ross, 1999) and having a parent who is employed (Keithly & Deseran, 1995; Leventhal et al., 2001; Purtell & McLoyd, 2011), factors that can be construed as forms of work socialization, have been found to predict paid employment among low-income adolescents. In the present study, we examined links between adolescent work and two family context variables—child domestic responsibilities and parents’ expectations about how much schooling they expect their child to complete. Based on our assumption that higher levels of domestic responsibility reflect greater parental demands for maturity, dependability, and self-reliance, and in keeping with Entwisle et al.’s (1999) finding, we hypothesized that adolescents whose parents expected them to assume more responsibility for household chores during childhood/early adolescence would have greater odds of being employed and longer periods of employment 5 years later. Second, we expected that adolescents whose parents had higher postsecondary educational expectations for them would have higher odds of employment, but work fewer hours because high work intensity might be seen as a hindrance to higher levels of educational attainment. Third, we hypothesized that greater parental investment in adolescent’s acquisition of middle-class cultural capital (e.g., taking them to museums) would be concurrently associated with increased odds of employment, but lower work intensity. The latter prediction is based on the assumption that greater parental investment in adolescents’ middle-class cultural capital signals higher parental educational expectations for the adolescent.
Community Context
Relationships and networks that are capable of transmitting valuable resources (e.g., information and assistance) constitute social capital (Lin, 2001). Networks of people have stores of resources at their disposal, and these resources are exchanged through social interactions. Social capital that inheres in relationships can be used to generate other forms of capital, including human and cultural capital (Ream, 2005). Just as having an employed parent increases the likelihood of adolescent work (Keithly & Deseran, 1995; Leventhal et al., 2001; Purtell & McLoyd, 2011), there is reason to believe that employed individuals who are not family members but part of a family’s social network can also increase the odds of adolescent work. Multiple qualitative studies (Blustein et al., 2002; MacLeod, 2009; Newman, 1999) have documented that knowing adults with connections to specific employment opportunities is the most common route to obtaining a job for low-SES youth. Although personal connections are generally advantageous at all levels of the labor market, they are particularly useful in the low-wage service market where there is often a large supply of workers who are qualified for the jobs in question (Newman, 1999). In the present study, we hypothesized a positive concurrent relationship between employment and adolescents’ social capital among nonparental adults, such that adolescents who had close relations with more extrafamilial adults (e.g., teachers, mentors) and more extended family adults (e.g., aunts, uncles) would have greater odds of employment.
The level of economic and social vitality of the neighborhood or local geographic area where adolescents reside also appears to influence the odds of adolescent work. Using detailed geo-coded data, Weinberg, Reagan, and Yankow (2004) found that male teenagers who resided in economically and socially disadvantaged neighborhoods worked less than those in areas with more economic resources and higher levels of social organization. Likewise, spatial isolation from non–poor households is associated with decreased likelihood that urban minority youth will work during the late adolescent years, a relationship that may be due to transportation difficulties and reduced contact with employed adults who can provide information and connections needed to obtain work (Gardecki, 2001; O’Regan & Quigley, 1996). Other research has shown that residing in high poverty neighborhoods, regardless of urbanicity, decreases youth employment (Gardecki, 2001). In the present study, we used a measure of neighborhood social organization in the absence of indicators of the economic well-being of the neighborhoods where families resided. We predicted a positive concurrent relationship between neighborhood social organization and adolescents’ odds of being employed.
Race as a Moderator of Psychological Predictors of Adolescent Work
The second major goal of the present study was to assess whether links between adolescent employment and psychological factors (i.e., self-reliance, behavioral problems, expectations of 4-year college completion) were conditional on race. Low SES adolescents and Black adolescents, regardless of socioeconomic background, are less likely to be employed than middle-class White youth (Gardecki, 2001; Lerman, 2000). They may experience more difficulty obtaining and holding jobs than middle-class White youth because of discrimination, spatial isolation, a more limited and competitive local job market, and social networks that have proportionately fewer individuals who can connect them to employment opportunities (Bachman et al., 2013; Bertrand & Mullainathan, 2004; Gardecki, 2001; Hwang & Domina, 2017; Weinberg et al., 2004). A more stringent process of finding and maintaining employment may mean that low-SES youth and Black youth who decide to work, get hired, and keep jobs for longer periods of time may have higher levels of psychosocial assets (e.g., self-reliance, postsecondary educational aspirations, self-control) and lower levels of problem behaviors than middle-class White youth who are employed (Bachman et al., 2013; Hwang & Domina, 2017). Following this line of reasoning, we predicted a stronger positive relationship between work status (and duration and intensity of employment) and psychological assets (self-reliance, expectations of 4-year college completion) among Black adolescents than White adolescents. Conversely, we hypothesized a stronger negative relationship between work status (and duration and intensity of employment) and behavioral problems among Black adolescents than their White counterparts.
Method
Participants
Data for this study were drawn from the Panel Study of Income Dynamics (PSID)–Child Development Supplement (CDS). The PSID is a world renowned panel study that began in 1968 with a nationally representative sample of more than 18,000 adults living in 5,000 families in the United States. Data from this original sample and their children (and grandchildren) have been collected continually during the ensuing decades, making it the longest running household panel study in the United States.
In 1997, the CDS was added to the PSID with the purpose of obtaining more detailed information about how economic and social factors are related to developmental outcomes among children. The sample was drawn from 1997 PSID households with children below the age of 13 years. While 2,705 households met these criteria, a total of 2,394 households with a total of 3,586 children were included, resulting in a response rate of 88.2%. In addition to the first wave, subsequent waves were collected in 2002 and 2007, resulting in response rates of 91% and 90%, respectively. Data are comprehensive and include information sourced from child self-report, parent self-report, teacher report, child assessment, and child diaries, all of which are tied to the family-level data collected in the household surveys.
The initial sample used to assess adolescent employment status during the school year is limited to those children who participated both in 1997 (Wave 1) and 2002 (Wave 2) so that both childhood and adolescent variables could be concurrently examined as predictors of adolescent outcomes. Given our theoretical rationales and corresponding goal to examine interactions by race, the sample is further limited to those participants who identified as being either Black or White. Other racial groups were omitted because dummy-coding them as individual groups would complicate efforts to explore interactions, whereas collapsing them together with Black participants into a “non-White” category would limit any meaningful interpretation of the interactions. Finally, the sample is further limited to those participants who were 12 to 19 years of age and attending school in 2002 (Wave 2; N = 1,103; 50.5% female, 48.7% Black; Wave 2: Mage = 15.33 years; Mincome/needs = 3.1; Meconomic strain = 1.65). Because age was calculated at the time of interview, and interviews were conducted over several months during the school year, the correspondence in ages between Wave 1 (6-13) and Wave 2 (12-19) is 6 years instead of 5 years for some individuals who had birthdays near their interview dates. Due to the fact that questions about duration and intensity of employment were only asked of participants who indicated that they were employed, analyses examining duration and intensity of employment were conducted on the subset of the sample that indicated that they were currently employed during the school year (n = 190; 50.5% female, 38.9% Black, Wave 2: Mage = 16.64 years).
Measures
Adolescent employment
Three measures of adolescent employment were assessed as three distinct outcomes. The first was current employment status during the school year, which was ascertained by asking students whether they were currently employed (interviews were conducted only during the school year; 1 = employed, 0 = not employed). Employment duration was assessed by asking students how long they had been at their current job and standardizing responses into number of weeks. Employment intensity was assessed by two questions asking students (a) how frequently they worked each month in the past year and (2) about how many hours they worked each time. Responses were standardized to reflect the average number of hours per week the adolescent worked at his or her current job. All employment measures were assessed in 2002 (Wave 2) and were based on adolescent self-reports.
Predictors
Employing an ecological systems framework, we examined three levels of influence (i.e., psychological, family, community) as predictors of adolescent employment.
Psychological factors
Adolescents’ level of self-reliance during childhood is a single parent-reported item from 1997 (Wave 1) asking whether “child tries to do things for (himself or herself), is self-reliant.” Responses ranged from 1 to 5, where 1 = not at all like my child and 5 = totally like child. Higher scores correspond to greater child self-reliance.
Adolescents’ level of behavioral problems during childhood is a summed composite score of 30 parent-reported items from 1997 (Wave 1) asking how often the child exhibits various negative internalizing and externalizing behavioral problems. Example items include “child has sudden changes in mood or feeling” and “child breaks things on purpose or deliberately destroys (his or her) own or another’s things.” Response items range from 1 to 3, where 1 = not true, 2 = sometimes true, and 3 = often true. Prior to summing these scores, not true was recoded as 0 and both sometimes true and often true were collapsed into a single category coded as 1. Higher scores correspond to higher levels of behavioral problems.
Adolescents’ postsecondary educational expectations is a single adolescent-reported item from 2002 (Wave 2) asking “what do you think are the chances that you will graduate from a 4-year college?” Responses ranged from 1 to 5, where 1 = no chance and 5 = it will happen. Higher scores correspond to higher educational expectations.
Family context
Adolescents’ chores/domestic responsibilities during childhood is a mean composite score of five parent-reported items from 1997 (Wave 1, α = .75) asking how often the child is expected to complete various chores. Example items include “clean (his or her) own room” and “do routine chores such as mow the lawn, help with dinner, wash dishes, etc.” Responses ranged from 1 to 5 where 1 = almost never and 5 = almost always. Higher scores indicate greater domestic responsibility.
Parents’ expectations for their adolescent’s educational attainment is a single parent-reported item from 2002 (Wave 2) asking “how much schooling do you expect that your child will complete?” Responses ranged from 1 to 8, where 1 = 11th grade or less and 8 = MD, law, PhD, or other doctoral degree. Higher scores indicate higher parental expectations.
Adolescents’ middle-class cultural socialization is a mean composite score of two parent-reported items from 2002 (Wave 2) asking how often a family member has taken the adolescent to a musical or theatrical performance in the past year. Response items range from 1 to 5, where 1 = never and 5 = more than once a month. Higher scores correspond to higher levels of middle-class cultural socialization.
Community context
Adolescents’ social capital is a sum composite score of 16 adolescent-reported items from 2002 (Wave 2) asking whether they are close to teachers and other nonparent adults such as aunts, uncles, mentors, or a friend’s parent. Response items were dichotomized where 0 = “no” and 1 = “yes.” Higher scores reflect higher levels of adolescent social capital.
Neighborhood social organization is a mean composite score of seven parent-reported items from 2002 (Wave 2; α = .77). Example items include the following: “How difficult is it for you to tell a stranger in your neighborhood from someone who is a resident?” “How likely is it that a neighbor would do something if your kids were getting into trouble?” and “How would you rate your neighborhood as a place to raise children?” Response items ranged from 1 to 3 (not at all difficult to very difficult), 1 to 4 (very unlikely to very likely), or 1 to 5 (excellent to poor), respectively. Prior to being averaged, the items were standardized to 1 to 3 by collapsing categories. Higher scores correspond to higher levels of neighborhood social organization.
Demographic controls
Demographic variables used as controls included child gender (0 = female, 1 = male), race (0 = Black, 1 = White), and adolescent age in years in 2002. Two economic factors were also included as controls: (a) a composite of the average family income-to-needs ratio for the 3 years prior to the 1997 CDS (i.e., 1994, 1995, 1996; Wave 1) and (b) family economic strain. Because the average family income-to-needs ratio, in its original form, violated assumptions of normality (skewness = 7.58; kurtosis = 94.33), it was transformed by computing its logarithmic conversion, resulting in acceptable skewness (–0.41) and kurtosis (0.52).
The measure of family economic strain is a summed composite score of 16 parent-reported items from 2002 (Wave 2) asking whether in the past 12 months the parent or family had certain experiences or taken various actions as the result of economic problems (e.g., “had a creditor call or come to see you to demand payment,” “moved in with other people,” “borrowed money from friends or relatives”). Response items were dichotomized where 0 = “no” and 1 = “yes.” Higher scores reflect greater economic strain.
Analysis Plan
Three multiple regression models were run for each of the three employment outcomes using Stata 14. To correct for nonindependence between siblings on variables assessed at the household level (i.e., family economic strain, family income-to-needs ratio), clustering was employed. In addition, we applied multiple imputation with chained iterations to account for missing data, which was less than 5% on all predictor variables. To test whether psychological variables may matter more for Black adolescents’ employment, we explored interaction effects between race and each of the three psychological factors (self-reliance, behavioral problems, and postsecondary educational expectations) on employment status, duration of employment, and intensity of employment, including them simultaneously in each model.
Results
Table 1 presents basic descriptive data for all nondemographic variables. Among the sample of 1,103 participants, 190, or just over 17%, reported that they were employed during the school year. Notably, the average number of weeks in one’s current job varied substantially, from having recently started to having been working for 8 years with the same person or company. Likewise, there was substantial variation in the number of hours worked each week, with some adolescents working as few as 30 minutes and others working up to 56 hours. Furthermore, 36% of adolescents reported working more than 20 hours per week.
Descriptive Data for Nondemographic Variables.
Correlations between all study variables are shown in Table 2. Although there are several significant correlations, there is no evidence of multicollinearity, as none of the correlations exceeded .8. Notable relationships include the fact that race is positively correlated with employment status, but negatively correlated with employment intensity. Parent postsecondary expectations are moderately correlated with adolescent postsecondary expectations, indicating congruence between these variables that supports their respective validity. Finally, the psychological variables (self-reliance, behavioral problems, postsecondary educational expectations) are all correlated with each other in meaningful ways (i.e., behavioral problems are negatively correlated with self-reliance and educational expectations). However, none are associated with employment outcomes in the absence of controlling for other variables.
Correlations Between All Study Variables (N = 1,103).
Note. Employment duration and employment intensity, n = 190.
p ⩽ .05. **p ⩽ .01. ***p ⩽ .001.
Table 3 shows the results of each of the multiple regression models, all of which included the same predictor variables. Among these full models, only Models 1 and 3 were significant, (Model 1: Employment status F(16, 5.3e6), p < .001 and Model 3: Employment intensity F(16, 2.8e06), p < .001). Although the full version of Model 2 was not significant, employment duration F(16, 6.4e6), p = .13, it included significant predictors that, when entered alone, resulted in a significant model (not shown), suggesting that the nonsignificance of the full model is due to the additional variables that fail to explain variance. Nonetheless, we chose to keep the additional variables in the model for the purpose of comparison across outcomes.
Predictors of Employment Status, Duration, and Intensity in 2002.
p ⩽ .05. **p ⩽ .01. ***p ⩽ .001.
Instead of presenting our results separately for each dimension of employment (status, duration, and intensity), results are organized by demographic, psychological, family, and community variables across the three dimensions of employment. Foregrounding these four categories of variables is consistent with the primary foci of the present study and aligns well with key themes in the research literature.
Demographic Controls
Among the demographic variables, only race and age predicted adolescent employment outcomes. White students were 2.5 times more likely than their Black counterparts to be employed. However, among those employed, race did not predict employment duration or intensity. Age had a positive association with both employment status and employment intensity. A 1 year increase in age was associated with a 1.79 increase in odds of being employed and an increase in average hours worked of 2.44 hours. Gender, family income-to-needs ratio, and family economic strain were not associated with any of the employment outcomes.
Psychological Factors
Contrary to our expectations, most of the psychological factors in both childhood and adolescence did not predict adolescent employment status, duration, or intensity. The exception is that self-reliance significantly predicted employment duration, but the relationship was reverse of what we predicted. That is, greater self-reliance was associated with shorter, rather than longer duration of employment.
A significant Race × Self-Reliance interaction was found in the models predicting duration and intensity of employment, but not in the model predicting employment status. However, the nature of these interactions was contrary to our predictions. Specifically, among Black adolescents, greater self-reliance predicted shorter duration and lower intensity of employment. In contrast, among White adolescents, self-reliance was unrelated to employment duration, but positively associated with employment intensity.
In the models predicting employment status and duration of employment, the Race × Postsecondary Educational Expectation interactions were not significant. However, the interaction between race and postsecondary educational expectation was a significant predictor of employment intensity. Exploration of this interaction (see Figure 1) indicated that whereas White students’ expectation of college completion was negatively associated with employment intensity, Black students’ expectation of college completion was positively associated with employment intensity. The interaction between race and behavioral problems was not significant for any of the employment outcomes.
Interaction between race and postsecondary expectations on employment intensity.
Family Context
Consistent with our hypothesis, childhood domestic responsibilities was a significant predictor of employment status, such that a one-unit increase in childhood domestic responsibilities was associated with a 1.39 increase in odds of being employed. The association between middle-class cultural socialization and employment status was also statistically significant. However, contrary to our prediction, the relationship was negative such that a one-unit increase in cultural socialization was associated with a .72 decrease in odds of being employed. Parents’ educational expectations for the adolescent was unrelated to employment status, but predictive of employment duration, such that a one-unit increase in parent educational expectations was associated with an average decrease in employment duration of 9.55 weeks. None of the family context variables predicted employment intensity.
Community Context
Neither neighborhood social organization nor social capital was predictive of any of the employment outcomes.
Summary
In terms of family context, children who had more household responsibilities were more likely to be employed as adolescents. Adolescents who experienced more cultural socialization were less likely to be employed, and those whose parents had higher educational expectations for them worked shorter periods of time. Race operated as a moderator in two instances. Whereas greater self-reliance during childhood predicted lower intensity and shorter duration of employment among Black adolescents, it predicted higher intensity of employment among White adolescents. Expectation of college completion and employment intensity were positively associated among Black adolescents but negatively associated among White adolescents. Family context variables were unrelated to employment intensity. Community context variables did not account for significant variation among any of the indicators of employment.
Discussion
Previous research has found that Black adolescents are less likely to be employed than their White counterparts (e.g., Ihlanfeldt & Sjoquist, 1990; Raphael, 1998). The present study replicates this finding and suggests that this racial disparity is robust to the psychological, family, and community variables we examined. Local economic opportunity is likely a factor that contributes to this racial disparity, although this does not rule out psychological and motivational factors that may differentially select Black and White adolescents into employment. Our finding that older adolescents are more likely to be employed and work longer hours per week than younger adolescents is congruent with previous research (e.g., Yamoor & Mortimer, 1990) and intuitive given developmental processes that link increasing responsibility with maturity.
Previous research indicates that autonomy (Purtell & McLoyd, 2011), behavioral problems (Caspi et al., 1998; Daly et al., 2015), and adolescent academic expectations (Entwisle et al., 2005; Lee & Staff, 2007; Steinberg et al., 1993) are predictive of later employment outcomes. In the present study, we examined similar psychological variables and to our surprise, they tended to be unrelated to work status, duration, or intensity. Although self-reliance was predictive, it was not in the expected direction, as higher self-reliance was associated with lower duration of employment. We believe that this latter finding may have to do with limitations in the measurement of duration. Our measure assessed how long the adolescent had worked in their current position. As such, it fails to capture how long the adolescent had maintained employment across job positions and across a substantial period of time (e.g., a 1-year period). Adolescents with higher levels of self-reliance may be more likely to leave jobs to take advantage of better employment opportunities without necessarily having gaps in employment. Our measure of employment duration would not capture such instances. Future studies should examine not only duration in current position but also the duration of employment and number of positions held over the course of an established period of time. A more stable measure of this nature would permit a more stringent test of our hypotheses about precursors of employment duration and how they interact with race.
In the present study, behavior problems during childhood did not predict employment outcomes. Given that our measure of behavioral problems was based on parent reports, and assessed during a developmental stage (i.e., early to late childhood) when proximity to parents is higher and thus under more scrutiny compared with adolescence, it is possible that there is incongruence between parent perceptions of these variables in childhood and employer perceptions in adolescence. In short, problem behaviors at home during childhood may not manifest themselves in ways that are apparent to employers, particularly during the hiring stage.
A main effect was not found for adolescent postsecondary expectations, but this variable interacted with race in the prediction of employment intensity. Black students’ expectation of college completion was positively associated with employment intensity, whereas White students’ expectation was negatively associated with employment intensity. This finding is important to untangling the relationship between academic aspirations and work intensity, particularly in light of Marsh and colleagues’ (Marsh, 1991; Marsh & Kleitman, 2005; Nagengast et al., 2014) findings that the motivation for working intensively is more predictive of educational aspirations and college attendance, rather than intensity per se. Our finding suggests the possibility of race differences in motivation to work during adolescence, such that Black students may tend to view intensive work and college attendance/completion as mutually congruent with upward social mobility, whereas White students may be more inclined to view them as mutually exclusive. This seems highly plausible given race differences in income and wealth that may decrease Black parents’ ability to pay college expenses and increase pressure on Black adolescents to work and save money for college (Oliver & Shapiro, 2006; Zhan & Sherraden, 2011). Future research should explore whether there are systematic race differences in adolescents’ motivation to work intensively, provide additional tests of whether race moderates the association between various dimensions of adolescent employment and postsecondary educational goals, and examine the contribution of wealth in these relationships.
On conceptual grounds and in line with prior research findings (Entwisle et al., 1999), we hypothesized a positive relationship between domestic responsibilities during childhood and both employment status and duration, but not intensity during adolescence. We found that more domestic responsibilities was associated with higher odds of being employed, but was unrelated to employment duration. This suggests a socialization effect of early participation in household labor on later paid labor. This socialization might lead to improved labor market outcomes by (a) increasing motivation to enter the labor market in the first place and/or (b) increasing marketable “soft skills” (e.g., dependability, initiative, sense of responsibility) that employers view as valuable, thus increasing odds of being hired, compared with other equally motivated applicants. Our understanding of these issues would be advanced by studies that examine not only current employment, but also adolescents’ job search behavior and self-described desire to work as they relate to domestic responsibilities and other parental socialization processes during childhood.
Contrary to our prediction, parental postsecondary expectations were negatively associated with employment duration. Parents who have high expectations for their adolescent’s college attendance/completion may limit their adolescent’s commitment to work to enable the adolescent to maintain or increase their GPA or participate in extracurricular activities. In addition, parents with high postsecondary educational expectations may be more likely to have accumulated college savings, thereby reducing adolescents’ need to remain employed for extended periods. Parent and adolescent expectations of college completion were positively correlated in the present study, but little is known about the level of agreement between parents and adolescents about the reasons for employment and the optimal intensity of employment. Other interesting questions that warrant further study concern the nature of parent–adolescent negotiations around issues pertaining to employment, how adolescent employment influences the nature and quality of parent–adolescent relations, and the extent to which these processes vary by social class and race.
More middle-class cultural socialization was associated with lower odds of being employed, but was unrelated to employment intensity. Again, it may be that parents who are culturally socializing their children for college readiness may also be limiting their employment so as to leave time for other activities seen as more valuable precursors to competitive college admissions, such as homework or extracurricular activities. Future research should more fully explore how parental control, autonomy, and a variety of socialization behaviors—particularly around postsecondary readiness—may impact adolescents’ pursuit of employment during the school year, and how these connections vary as a function of race, SES, and the interaction of race and SES.
Adolescent social capital was unrelated to any of the dimensions of employment examined in the present study. This may be due, in part, to the fact that we measured the quantity of adolescents’ close connections without addressing the quality of those connections or the characteristics of individuals with whom adolescents had close relations (i.e., whether or not they were employed). The lack of an association between neighborhood social organization and adolescent employment status may reflect the insufficiency of our measure as a proxy for neighborhood-level economic disadvantage.
Although our hypotheses were informed by prior research and seemed conceptually well-grounded, most were not supported or were contradicted by the findings. We surmised a number of plausible explanations for these incongruities. In addition to the previously mentioned limitations of some of the measures, other limitations include the comparatively small sample size used in the employment intensity and duration models, as well as our inability to cross-lag childhood predictors at multiple points in time prior to adolescence, which would be a more ideal method for confirming selection effects.
We believe that our hypotheses about race as a moderator of psychological predictors of adolescent work warrant further testing in longitudinal research with more reliable measures of employment and stronger, more extensive measures of psychological characteristics. Such work will constitute an important step toward resolving the question of differential selection into employment as a contributing factor to race differences in the outcomes of employment. More generally, future research should continue to explore adolescent employment through the lens of an ecological systems model that aims to parse out how the environment in conjunction with personal dispositions influences employment and, in turn, how adolescent employment predicts early adult educational and employment trajectories.
Footnotes
Authors’ Note
Samantha K. Hallman, Department of Psychology, University of Michigan. Samantha Hallman is now at the Department of Social Work, Madonna University.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The collection of data used in this study was partly supported by the National Institutes of Health under grant R01 HD069609 and the National Science Foundation under award 1157698.