Abstract
Student absenteeism in the earliest years of elementary school has been linked to a range of negative outcomes. Though the literature has examined numerous factors that are associated with children missing school, the role of teachers—especially at the early elementary level—has not been well understood. Given that students spend the majority of their time in elementary school with one teacher in a single classroom, a large component of early elementary school is underexplored, and our understanding of absenteeism remains incomplete. In this vein, we looked into whether having novice versus more experienced teachers was linked to the frequency of student absenteeism in the earliest years of school. Using a national dataset including repeated observations of students in grades K-2, we found that novice teachers have students with fewer absences and a lower probability of being chronically absent. There were no differences in the outcomes of novice teachers as a function of child nor classroom characteristics. Implications for policy and practice are discussed.
As is common in other professions (Ng & Feldman, 2010), effectiveness in teaching is often contextualized by the number of years on the job (i.e., what we refer to as “experience” in this study), with a keen focus on effectiveness during the novice years of a teacher’s career (Podolsky et al., 2019). This heightened attention to the early years of teaching—though well established in the teacher effectiveness literature—is based primarily on studies measuring teacher efficacy in terms of student achievement. Although an important marker of student success, there are other key outcomes that are important and predictive of not only good teaching, but also students’ long-term life success (e.g., Ansari et al., 2020). Reflecting the above, researchers have begun to expand our understanding of teacher years-on-the-job to measure the effects of teachers on a broader set of outcomes, such as growth mindset and executive function, which are outcomes our education system also targets (e.g., Attendance Works, 2016; Kraft, 2019).
With regards to broadening our thinking as to how teachers with different years of experience might differentially influence student outcomes, one area that has received less attention is student absenteeism. Yet, student absenteeism is a critical measure of school success (Gottfried et al., 2019): Higher rates of absenteeism from school have been linked, causally, to a host of adverse educational outcomes not only in the year of missing school but also future years (e.g., Gottfried, 2011a; Gottfried et al., 2019; Coelho et al., 2015; Liu et al., 2021). Importantly, detrimental effects of absenteeism have been detected as early as in elementary school (Anderson & Romm, 2020; Gottfried, 2009, 2010, 2014; Gershenson et al., 2017). During these earliest years of schooling, absenteeism has also been linked to lower test performance of other students in the classroom (Gottfried, 2011b), thereby showing how there is a permeating effect of individual absenteeism in the context of the classroom. And despite the perception that absenteeism is only worthwhile of attention and remediation in adolescence (Gottfried et al., 2019; Robinson et al., 2018), it has been shown on a national scale that children in the earliest years of elementary school, including kindergarten and first grade, have some of the highest rates of absenteeism—rates not witnessed again until middle and high school (Balfanz & Byrnes, 2012).
Given the significant consequences associated with absenteeism, there has been a great deal of research that has considered why students are absent. Although high rates of absenteeism among young children has been shown to be indicative that some parents are disconnected from their children’s schooling (Catsambis & Beveridge, 2001; Fan & Chen, 2001; Jeynes, 2003; McNeal, 1999), it might not necessarily signal family disengagement. In many cases, parents are often unaware of or underestimate how many days their children are absent (Rogers & Feller, 2018). This lack of awareness may be exacerbated for low-income families, where parental engagement and awareness are even lower (Abrams & Gibbs, 2002; Epstein, 2001; O’Connor, 2001). Research has also shown that children benefit when a stronger link is made between parents and schools (Sheldon, 2007), particularly in low-income families. Thus, when determining how to allocate resources to reduce absenteeism, it is crucial to understand what school-factors, programs, and processes might be linked to higher and lower absenteeism in order to increase connections to school as well as engagement (Gottfried et al., 2019).
Within this area, many studies have underscored the importance of policy-manipulable school factors that might affect absences, such as schoolwide partnerships with parents and/or the community (e.g., Childs & Grooms, 2018; Childs & Lofton, 2021), as well as programs that are designed to foster school connections between schools, children, and parents (Mac Iver & Sheldon, 2019). In this study, we also examine a key school factor that might improve engagement and connections and ultimately lower absenteeism. Namely, our study focuses on teachers. We do so given emerging research which suggests that teachers can play an important role in student attendance. For instance, Gottfried et al. (2022) found that high school students had fewer unexcused absences when teachers shared the same race/ethnicity. Similarly, studies have shown that teachers with higher value-added scores reduced high school students’ absences (Jackson, 2018; Liu & Loeb, 2019), and Gershenson (2016) found that teachers with higher levels of measured effectiveness reduced third, fourth and fifth graders’ absences. Yet, minimal attention has been paid to how experience on the job might affect students’ absences, especially the distinction of novice versus more experienced teachers. Thus, we consider how novice teachers, relative to more experienced teachers, might link to student absenteeism in early elementary school. Specifically, we ask:
To what extent does student absenteeism differ for students taught by novice teachers relative to those taught by more experienced teachers?
Do the findings vary for students who may experience higher risks of absenteeism (e.g., students with disabilities) or for students with different types of teachers or in different types of classrooms?
To answer these questions, we drew on data from the Early Childhood Longitudinal Study- Kindergarten Class of 2010 to 2011 and focus on kindergarten through second grade given that absenteeism in elementary school is at its highest points in these grades (Balfanz & Byrnes, 2012; Ehrlich et al., 2018). Our study demonstrates that students taught by novice teachers were absent less frequently than those taught by more experienced teachers, which runs contrary to the common (mis)perceptions that novice teachers are less effective for student success.
When taken together, this study helps provide a richer understanding into the classroom contextual factors that might impede or support school attendance during the early years of school that are plagued with high rates of absenteeism. In addition, this research also comes at a time when there is growing interest in and action to include non-academic outcomes, like absenteeism, in the design of accountability systems (Gottfried et al., 2019; Marsh et al., 2017). Our work generates valuable evidence to inform these discussions. Finally, many of the factors identified as impactful on student absenteeism are often beyond school context, such as public transportation, health, and family/household characteristics (Gottfried, 2015; Kerr et al., 2012; Morrissey et al., 2014; Robinson et al., 2018; Weaver, 2002). Yet, if having a novice teacher versus more experience teacher links to student absenteeism during the earliest years of education, this provides an opportunity to better understand how unexplored classroom-level factors might be linked to supporting or inhibiting positive school-going behaviors.
Background
Absenteeism in the Earliest Elementary Grades
Approximately 25% of all kindergartners in the United States are missing large amounts of school (often 10% or more of the school year), and this trajectory of absenteeism carries forward throughout children’s educational careers (Romero & Lee, 2007). Given this, a robust literature has now been established around the effects of absenteeism on a host of outcomes, ranging from lower academic outcomes to more behavioral issues (e.g., Gottfried, 2014; Chang & Romero, 2008; Gershenson et al., 2017). As early as in kindergarten, there are links between absenteeism and lower school performance. For example, Romero and Lee (2007) found that higher absenteeism is associated with lower academic performance, particularly for students from low-income homes. Others have also examined and identified differences in absenteeism and/or its effects by subgroup, such as income, English language learners, or those with disabilities (e.g., Anderson & Romm, 2020; Garcia & Weiss, 2018; Gee, 2018; Jacob & Lovett, 2017; Nauer et al., 2014), which motivates our present work to examine group differences through our second research question.
Beyond academic achievement, using a nationally-representative sample of kindergarten students, Gottfried (2014) found that high rates of absenteeism were associated not only with lower math and reading achievement, but also with less optimal socio-emotional development, including engagement with academic content in the classroom and social engagement. Another study using the same dataset but focused across the whole elementary school continuum found that the associations between absenteeism and student outcomes did not generally vary from grade to grade, but that compounded years of absenteeism was particularly detrimental across student outcomes (Ansari & Gottfried, 2021).
Novice Versus Experienced Teachers
Substantial literature has examined the links between teacher experience and student outcomes (Henry et al., 2011; Podolsky et al., 2019). Much of this research conceptualizes teacher experience in a dichotomous fashion in terms of novice and experienced teachers (e.g., Ingersoll & Merrill, 2012). Although there is variation in definitions of novice and experienced, novice teachers are often defined as having experience somewhere in the range of 0 to 3, 0 to 4, or 0 to 5 years—though as we describe in our analyses, these terms do take on different definitions, and we will explore these. As we will also detail below, this is in large part because the literature generally finds substantial gains to experience in the earliest years of teachers’ tenures (e.g., Kini & Podolsky, 2016). Beyond gains, the prominence of the novice versus experienced terminology distinction is also commonly used because novice teachers make up a large portion of the teaching workforce. Indeed, the modal category of teaching experience in the U.S. is 1 year, and large proportions of these teachers leave the workforce within a few years (Ingersoll & Merrill, 2012). This reality has focused much attention on novice teachers and ways to support their development in terms of effectiveness. Given the prominence of focus on these two distinct groups of teachers, we focus our terminology on novice and experienced teachers.
Summarizing a review of 30 studies, Kini and Podolsky (2016) wrote, “Teaching experience is positively associated with student achievement gains throughout a teacher’s career. Gains in teacher effectiveness associated with experience are most steep in teachers’ initial years, but continue to be significant as teachers reach the second, and often third, decades of their careers” (p. 1). Yet, in all studies, gains to experience were steepest when teachers were novice. An analysis from Henry et al. (2011) drew on data from elementary school teachers in North Carolina and found that gains to experience were steepest between the first and second year of teaching and diminished thereafter. The magnitude of impacts found in studies like these were quite large. For example, Wiswall’s (2013) analysis of elementary school classrooms in North Carolina found that the measured quality of teachers with 30 years of experience, in terms of math achievement, was one standard deviation higher than beginning novice teachers. Teachers with 30 years of experience also performed 0.75 standard deviations higher than novice teachers with 5 years of experience—demonstrating the continuation of gains to experience even after the first few years (Papay & Kraft, 2015).
Although the literature focused on years of experience and student outcomes has primarily focused on the outcome of student achievement, there is a limited body of evidence to suggest that years of experience is also associated with important non-academic outcomes, such as student behavior (e.g., Ladd & Sorensen, 2017). This research has come in response to a recent shift in the teacher effectiveness literature demonstrating that teachers play a significant role in promoting a broad range of outcomes, such as growth mindset (e.g., Kraft, 2019).
Novice Teachers and Student Absenteeism
In terms of absenteeism, limited and inconclusive literature has examined the link between teachers’ years of experience and absenteeism. Ladd and Sorensen (2017) drew on a sample of middle school students in North Carolina from 2007 to 2011 to examine links between years of teaching and student absences, in addition to disciplinary offenses, time spent reading for pleasure, and time spent completing homework. The authors found that returns to experience were largest for student absenteeism out of all of the outcomes they had examined. Gershenson (2016) also examined experience and absenteeism in North Carolina in addition to a longitudinal dataset of kindergartners and their teachers, beginning in 1998 to 1999. Across both studies, he found that years of teaching was linked to fewer absences. Arguably the most causal study however is Tran and Gershenson (2021) who relied on experimental estimates of the way that teachers were assigned to students to examine whether years of experience impacted students’ absenteeism in early elementary school. The authors did not find a statistically-significant effect of years of experience on student absences.
Certainly the mixed findings in the small body of literature call for additional studies to help understand the effects of years of teaching. In addition, other ways of conceptualizing years of teaching beyond simply the number of years in the profession merits being explored, as the teaching field itself differentiates teachers not simply by number of years in the profession but by new versus experienced teachers. In this direction, our study builds upon previous work by examining how having novice versus experienced teachers might play out during an earlier peak of absenteeism—early elementary school—which, as mentioned, sees absences as high as in middle school. This critical time period of education is of concern given that school going routines and habits are easier to modify in the early years, which in turn, set the stage for more regular school attendance in the later years (Gottfried et al., 2019). Accordingly, our study builds upon this earlier work in two important ways. First, we provide generalizable, nationally-representative estimates of the association between having novice versus experienced teachers and student absenteeism—a distinction in teaching that is relevant and pervasive in the teacher literature. Second, we focus our analysis on the earliest grades of schooling (i.e., K-2), where absenteeism rates are as high as in early adolescence, yet potentially more modifiable given students’ early age where patterns have not been fully solidified.
The broader existing literature on teacher experience and several of the previous experience-absenteeism studies showing its positive correlation with student attendance might lead one to hypothesize that experienced teachers will be more effective in reducing student absenteeism than novice teachers in early elementary school. This hypothesis stems from the fact that teachers with more experience may engage more effectively with parents and students around the importance of regular school attendance given that they have had more time to develop effective strategies to navigating these issues. The above is particularly relevant in the early elementary school years when parents are more involved in ensuring regular school attendance as opposed to the older grades when students have greater autonomy.
But one could also put forward a second hypothesis where having novice teachers could be linked to lower absences for children in the earliest grades of school. For example, novice teachers tend to be younger on average, which implies that the age difference might be small between them and the parents of children in early elementary grades. Parents play a critical role in school attendance (Chang & Romero, 2008; Rogers & Feller, 2018; Sheldon, 2007). Thus, a smaller age gap between parent and teacher might work to novice teachers’ benefit when it comes to addressing absenteeism: Novice teachers can potentially develop stronger relationships with parents of young schoolchildren given this smaller age gap. There is a wealth of literature suggesting that factors of similarity, such as age, predicts stronger interpersonal connections (e.g., Reagans, 2011). While not examined in schools per se, a better understanding between individuals in formal settings often emerges with a closer age similarity (Perry et al., 1999). Under this hypothesis, then, this might stand in contrast, for instance, to the age difference between novice teachers and parents with older children in middle or high school, where this age similarity might dissipate and so might connections between parents and teachers. Hence, in elementary school, newer teachers might influence absenteeism, where in middle or high school other factors (like having more experience) might come into play as students develop their own agency when it comes to getting to and from school.
Another potential hypothesis is that novice teachers are being prepared in a context that prioritizes new accountability structures and pedagogical approaches that more experienced teachers were less exposed to. For example, accountability policies in recent years have expanded beyond student test scores to include non-academic outcomes like attendance (Childs & Lofton, 2021). It could be that this policy environment and the messages it sends are most salient with novice teachers who are constructing their professional approaches, or they were prioritized in their pre-service preparation programs. Another potential influence is the emergence and prioritization of culturally relevant pedagogy in teacher preparation (Ladson-Billings, 2023; Muñiz, 2019). It could be that these approaches that have been shown to unlock greater engagement and buy-in from both students and families are more prevalent among more novice teachers. In short, teacher preparation and the contextual influences on teacher development shift over time, and there are a range of possible associated differences in practice that could translate to improving absenteeism among some cohorts relative to others.
There is of course a final possibility, such that no differences emerge in the absenteeism patterns of students as a function of having a novice versus more experienced teacher. Given the competing possibilities and limited literature in thinking about the role of novice teachers and students’ absences, we leave our study objectives as largely exploratory.
Method
Source of Data
For our analyses, we relied on data from the Early Childhood Longitudinal Study—Kindergarten Class of 2010 to 2011 (ECLS-K:2011). The National Center for Education Statistics at the U.S. Department of Education oversaw the creation and compilation of these data. Data were first collected on a nationally representative sample of children in kindergarten in the 2010 to 2011 school year, through both direct assessments as well as surveys given to parents, teachers, and school administrators. After kindergarten, the National Center for Education Statistics followed this sample through the end of elementary school. The ECLS-K used a three-stage stratified sampling strategy, in which geographic region represented the first sampling unit, public and private school represented the second sampling unit, and students stratified by race/ethnicity represented the third sampling unit. Hence, children in the ECLS-K database comes from a diversity of school types, socioeconomic levels, racial, and ethnic backgrounds.
Our outcome—absenteeism—was only an identical, repeated measure in the dataset from kindergarten through second grade because, in later grades, students have different teachers for different subjects who reported absenteeism, which makes the analysis fundamentally different than what is presented here with one teacher per grade and year. Therefore, examining early elementary school is a unique setting for examining absenteeism since students have the same teacher throughout the day and year. Therefore, our study relied on three observations per student, as they progress from kindergarten through second grade. The baseline kindergarten sample was approximately N = 18,170 students, where sample sizes are rounded to the nearest tens. We imputed all variables in our analyses for which there was missing data (Royston, 2004; Woods et al., 2023). Missing data on our focal predictor and outcome ranged from roughly 15% to 33%. We imputed 20 datasets for child observations for which nonzero weights (provided in the dataset by the National Center for Education Statistics and used in the analysis to follow) existed and included all focal variables of interest in our imputation procedures.
Absenteeism
Our dependent variables were absenteeism measures, and they can be found in Table 1. In the spring of kindergarten, first, and second grades, teachers reported the number of absences that the child had as part of their survey on the child’s performance in schools. The teacher had to select from a set of discrete answer choices on a 6-point Likert scale: 0, 1 to 4, 5 to 7, 8 to 10, 11 to 19, or 20 or more. Prior research using this dataset has recoded these six categories as 0 absences as 0, 20 or more absences as 20, and the midpoints for the choices of 1 to 4, 5 to 7, 8 to 10, and 11 to 19 (Gottfried, 2014, 2015, 2017; Ansari & Gottfried, 2021). In this study, we followed this coding. Additionally, per Gottfried (2014) and Gottfried & Kirksey (2021), we created an indicator for being chronically absent as having more than 10 absences. While chronic absenteeism often represents 10% or more of the school year (Jordan & Miler, 2017), most teachers in the dataset responded to these surveys in March of the year. Hence, 10 or more absences by March approximates 10% of the year. Note that there was only information on days absent, not days present.
Descriptive Statistics at Baseline in Kindergarten.
Note. Standard deviations are presented for non-binary measures.
p < .001. **p < .01. *p < .05 represent differences between groups.
Independent Variables
All independent measures can be found in Table 1, and all variables were identically collected and measured in kindergarten, first grade, and second grade. In the table, the variables are presented for the baseline kindergarten wave.
Novice Teachers
Table 1 is delineated by children in kindergarten who had novice teachers versus those who did not. In all waves of the data collection, teachers were asked to identify how many years they had been teaching. Guided by recent literature, novice teachers were defined as those with fewer than 3 years (e.g., Faez & Valeo, 2012; Garvis & Pendergast, 2010; Monte-Sano & Budano, 2013; Pogodzinski et al., 2013; Sutton, 2011). Note that the definition of “novice” does differ in the literature, and thus, we tested different definitions in our analysis presented below.
Control Variables
We relied on control measures that pertained to child, family, classroom, and teacher characteristics. Child characteristics included demographics, such as gender, race/ethnicity, parental rating of health (1 through 5, with one being the best health), whether English was not the primary language spoken at home, and disability status. We also included whether the child attended prekindergarten during the year before kindergarten, full-day kindergarten (vs. half-day kindergarten), and day-care before or after school during the kindergarten year. As shown in the first grouping in Table 1, children with and without novice teachers appeared to be relatively similar on these measures.
Next in the table are family measures. These include whether the child lived in a two-partner household and the number of siblings the child had at the time of the survey. Additionally, the child’s parents reported whether they chose the current school for their child (as opposed to it being assigned by default of residential zone) and whether the parents reported having chosen their house (or residence) for the specific school the child attended. For measures that reflect socioeconomic status, we included mother and/or father’s employment status as well as whether families were at or below the poverty threshold. Finally, given the particular importance of parental involvement in young children’s school attendance as opposed to other years of schooling (Gottfried, 2015), we included measures reported by teachers who were asked about a child’s parents’ involvement in schooling. We included a teacher’s rating for how well they perceived the parents to be involved (not, somewhat, very involved). In addition, we included indicators for a teacher’s report as to whether parents had participated in the following activities during the current school year (the teacher reported a yes/no): Attended regularly-scheduled conferences at school; attended parent/teacher informal meetings; returned teacher’s phone calls or emails; initiated contact with the teacher; volunteered in the teacher’s classroom on at school. Similar to the child characteristics, Table 1 indicates that families look similar across the novice and more experienced teacher delineations.
The third grouping of control variables in the table are classroom characteristics. These include size as well as percentages of classroom students who are girls, Black, Latinx, Asian, below grade level in reading, below grade level in math, disabled, and English language learners. Finally, the teacher rated classroom behavior (misbehaves very frequently, misbehaves frequently, misbehaves occasionally, behaves well, behaves exceptionally well)—we coded very frequent and frequent misbehavior as poor class behavior. As consistent with other measures in the table, classroom measures are similar when comparing classrooms with novice teachers to those classrooms with more experienced teachers.
The next set of control variables are teacher measures. These include race/ethnicity and whether the teacher shared the same race/ethnicity as the student, graduate degree attainment, and age. Perhaps unsurprisingly, novice teachers were less likely to hold a graduate degree and were younger.
Given differences in teacher characteristics by the novice and more experienced distinction, one concern might be that certain types of students might be differentially placed with novice versus more experienced teachers. That is, being paired with a novice teacher is a function of student characteristics that principals or other administrators might use to make classroom assignments. We address this concern in Table 2. In this table, we regressed having a novice teacher on a series of observable measures that might be used to match students; the linear probability models were run with school fixed effects (i.e., indicator variables for school, so as to compare students and teachers within the same school) and also included clustered errors at the school level. In the first set of columns, we relied on grades K, 1, and 2 and included child demographic measures as well as kindergarten entry assessments on reading, math, and behavior. Both the combined sample regressions as well as individual grade regressions do not provide evidence that having a novice teacher would be predicted by child demographic, academic, or behavioral measures.
Predicting Assignment to Novice Teachers.
Note. School-clustered errors are in parentheses.
In the second grouping of Table 3, we included lagged measures of child characteristics given that principals or parents might be assigning or requesting classrooms in the present year based on last year’s performance. This meant that that only grades 1 and 2 could be included in the regressions as there were no lagged measures in kindergarten. Again, the second block of regressions in the table do not indicate any evidence that having a novice teacher is linked to child demographics, academics, or behavior—neither in the current or previous year. This lack of significance on sorting children to teachers is consistent with prior research using ECLS-K data (Aizer, 2008; Neidell & Waldfogel, 2010; Wright, 2015). Hence, our evidence suggests that our results to follow are not due to sorting.
Effects of Novice Teachers on Absenteeism.
Note. All models include controls for month of assessment as well as US region. School-clustered errors are in parentheses.
p < .001. **p < .01. *p < .05.
Analysis Plan
To address our specific research questions, we began with a baseline regression model as follows:
where A is an absence measure for child i in classroom c in school s in year t. NOVICE is an indicator for having a novice teacher, as defined previously (though as mentioned we do test different definitions of novice further down in this study). We ran this model twice. First, we ran this model as presented in the equation for days absent. Second, we ran equation with a binary measure for being chronically absent. All control variables are those from Table 1 and include I for individual child characteristics, F for family characteristics, C for classroom characteristics, and T for teacher characteristics, including teacher ratings found at the end of Table 1. In all models presented in this study, the error term is clustered at the school. As consistent with Cho (2012), we tested an error structure that was clustered for individual instead of school. The absolute value of the standard errors did not change, however, in any meaningful way. Note all models included indicators for survey wave, month when the teacher completed the survey, and U.S. region.
A revision to this model is the inclusion of child fixed effects:
where δi represents indicators for deidentified child ID code. We relied on this model because it exploited within-child variation on having a novice teacher in 1 year but not in another year. Essentially, by looking within rather than across children, each child becomes their own comparison, and as such, only time varying covariates like having a novice teacher remained in the model. Out of the 54,520 student-year observations, approximately 20% of this longitudinal sample changed from having a novice to more experienced teacher, or vice versa, between years.
When using fixed effects in addition to multiple waves of the ECLS-K datasets to examine classroom or teacher factors, child fixed effects have been consistently supported as the preferred level of fixed effects (e.g., Aizer, 2008; Carbonaro & Maloney, 2019; Cho, 2012; Curran & Kitchin, 2019; Fletcher, 2010). Therefore, while we present baseline model results initially, after the first set of regressions, we only present the child fixed effects models. The motivation behind using an individual fixed effect is that by relying on a repeated observation of a child, we can use them as their own control. Hence, it is possible to better eschew unobserved forces and factors, such as individual motivation or principal sorting of children to a particular teacher. For the latter, if a principal assigns all of the children who have potential to be highly absent to a novice teacher, then the coefficient on β1 would be biased. Child fixed effects addresses this potential for within-school sorting (Fletcher, 2010; Wright, 2015), hence allowing us to better estimate the effect of novice teachers on child absences.
Results
Novice Teachers and Student Absenteeism
Our first research question addressed whether children with novice teachers had more or less absences compared with teachers who had been teaching for a longer amount of time. The findings for this question are found in Table 3. In this table, there are two groupings of results. In the first grouping, the outcome of days absent was regressed on an indicator for having a novice teacher as well as all control variables including indicators for survey wave. In the second, the outcome was a student being chronically absent, again regressed on the novice teacher indicator and control variables and including indicators for survey wave. In both groupings, column (1) represents equation (1) as the baseline model, and column (2) represents equation (2) as the child fixed effects model.
Across both days absent and chronic absenteeism outcomes, the findings suggest that children with novice teachers have fewer absences and are less likely to be chronically absent. This is evident by the negative, statistically-significant coefficients on the novice measure. Important to note is that the baseline model would have underestimated the size of the association between having a novice teacher and school absenteeism. Yet in contrast, when we ran our most-robust models including child fixed effects (i.e., columns 2 in both groupings), the findings suggest that children with novice teachers had fewer absences that year and a lower chance of being chronically absent compared to having teachers who had been teaching longer—for even fewer days or lower probabilities than what the baseline models would have suggested.
Variability in the Links Between Novice Teachers and Student Absenteeism
Our second research question looked into whether certain children or classroom settings were more or less likely to experience a “novice teacher effect.” For child characteristics, we examined health, English language learner, having a disability, and being in poverty—as these demographic characteristics are often associated with higher rates of absenteeism (Gottfried et al., 2019; Garcia & Weiss, 2018; Meng et al., 2012; Nauer et al., 2014). For teacher characteristics, we looked whether the teacher shared the same race or ethnicity as the student, whether the teacher had a MA degree or higher, and age. For classroom characteristics, we examined all of our classroom control variables from Table 1.
In Table 4, regression coefficients are presented for the interaction between our novice teacher indicator and the characteristic indicated running down the first row of the column. Each pair of coefficient and standard error represents the findings from a unique child fixed effects model, where this interaction was included along with the set of variables from Table 1. In other words, each cell is the interaction term from a child fixed effects model analogous to models (3) from Table 3. Overall, there is a lack of statistical significance throughout the table. This lack of moderation suggests that the link between absenteeism and having a novice teacher did not vary based on specific child, family involvement, teacher, or classroom characteristics. As a test of robustness, we also explored whether parents and teachers being close in age moderated the findings. The results were also not significant.
Differences by Students, Teachers and Classrooms.
Note. All models include controls for month of assessment as well as US region. School-clustered errors are in parentheses.
p < .001. **p < .01. *p < .05.
Sensitivity Checks
To test the robustness of our findings, we revisited our definition of novice teachers, as the literature might suggest different definitions of novice, anywhere from 0 to 5 years of teaching. For instance, some studies consider novice teachers as those with under 4 years of teaching (e.g., Glennie et al., 2016; Kidwell et al., 2021). Others consider novice teachers as those with under 5 years of teaching (e.g., Henry et al., 2012; Hogan et al., 2003). Table 5 presents the child fixed effects models from Table 3 for both outcomes. The first panel presents analyses using different definitions of novice: less than 4 years and less than 5 years. The findings are not statistically significant. Hence, this first robustness check provide additional confidence that our findings are specific to our study-wide definition of novice teachers, thereby suggesting a unique relationship that the newest-novice teachers have with the youngest elementary school students.
Tests of Sensitivity.
Note. All models are using student fixed effects. School-clustered errors are in parentheses.
p < .001. **p < .01. *p < .05.
In the second panel of Table 5, we sought to determine if our results were sensitive to the reference group of teachers compared to the novice group. In the first grouping, we included indicators for novice as well as for whether the teacher had more than novice but less than15 years of experience (“midrange” experience). Thus, the comparison (omitted) group were teachers with more than15 years of experience. In the second grouping in panel B of the table, we only included indicators for novice as well as for teachers with senior experience such that the midrange experienced teachers became the comparison. The results on novice, as shown in the table, were not sensitive to the comparison group. The coefficients on novice for both days absent and chronic absence models were similar to those presented in Table 3.
In the third panel of Table 5, we removed parent involvement measures in the child fixed effects regressions in the first column, as those were measured by the teacher in the spring at the same time as child absences. In the second column, we ran our regression removing all parental measures from the child fixed effects model. The results on having were not sensitive to the removal of these parental measures, thereby suggesting strength in our original findings as not being mediated by parental variables.
Finally, in the fourth panel (D), we explore a continuous measure of years of experience. First, we included both novice and a continuous measure of years of experience in the model. Essentially, years of experience here would be accounting for the fact that there is variation within the novice measure (i.e., some teachers have 1 year of experience whereas other have 2 years within this same binary category). In the second column, we removed our novice measure entirely from the model and only included a continuous measure of years of experience. Across the panel, years of experience was not a significant predictor of school absences. This finding corresponds to the experimental (and causal) findings in Tran and Gershenson (2021) and also underscores the importance of our study examining a more nuanced distinction in how teaching experience is measured—novice versus otherwise.
Additional Outcomes
The focus of our study was squarely on absenteeism. That said, we tested the robustness of our findings by examining several different outcomes that novice teachers might affect on the same sample of students—that is, achievement, socioemotional development, and executive function. We do so to examine whether our model is accurately estimating relationships with other outcomes in prior literature. To do so, we explored different outcomes that were provided in the ECLS-K dataset. First, we examined reading and math achievement scores, which were developed by NCES for the purposes of data collection. Second, we examined various measures of behavior, including self-control and externalizing behaviors. In the spring of each year, teachers would rate children based on a series of questions, on a 4-point Likert scale. From these questions, NCES created a self-control scale and externalizing behaviors scale. The self-control scale measures how well the child controls his/her temper, respects others’ property, accepts his/her peers’ ideas, and handles peer pressure. The externalizing behaviors scale measures the frequency with which a child argues, fights, gets angry, acts impulsively, and disturbs ongoing activities. Finally, NCES assessed working memory through the Numbers Reversed (NR) subtest of the Woodcock-Johnson III (Woodcock et al., 2001). NCES also assessed cognitive flexibility through the Dimensional Change Card Sort (DCCS; Zelazo, 2006).
Based on these measures, we reran our child fixed effects models. The findings are presented in Table 6. The findings are in the predicted direction based on what the literature has found for the effects of novice teachers on student outcomes. In Table 6, novice teachers were either negatively-linked to performance (i.e., reading) or do not have any statistically-significant link to student outcomes. These null results remained robust when we altered the definition of novice teachers. These results also remained robust when we tested additional socioemotional scales found in the ECLS-K dataset.
Additional Outcomes.
Note. All models are using student fixed effects. School-clustered errors are in parentheses.
p < .001. **p < .01. *p < .05.
Discussion
The extant literature has made clear that student absenteeism in the early elementary school years has a host of negative consequences for students’ academic achievement, executive function, and social-behavioral development, both in the short- and long-term (e.g., Ansari & Gottfried, 2020). As such, school absenteeism in the U.S. has been of growing concern for educators, policymakers, and practitioners alike (Ansari & Gottfried, 2018). Despite the extensive literature that has considered the causes (e.g., Morrissey et al., 2014; Van Eck et al., 2017) and consequences of student absenteeism (e.g., Gottfried, 2014; Ansari & Gottfried, 2021; Gershenson et al., 2017; Morrissey et al., 2014), to date, little attention has been paid to the ways in which those on the front lines, teachers, may influence students’ school attendance. The present investigation sought to address this gap in knowledge by bridging the literatures on teachers and student absenteeism in order to provide new insight regarding the extent to which having novice versus more experienced teachers is related to student absenteeism in the earliest years of school. In doing so, two important themes emerged.
First, novice teachers across the U.S. have children with fewer absences in the early elementary school years as compared with teachers with more years of experience. Although the estimates for the effects of novice teachers on the frequency of student absenteeism may appear small at first pass, the estimates reported herein are on par with the teacher experience literature (e.g., Wiswall, 2013) and other intervention and prevention efforts targeted at reducing student absenteeism (e.g., Robinson et al., 2018). Importantly, however, the positive effects of novice teachers were more sizable for those children on the cusp of being chronically absent, which represents a population of students at the center of intervention and prevention efforts (Ansari & Pianta, 2019). Given the limited research that has examined the link between teacher characteristics, including teacher experience, and absenteeism, our work addresses an important gap in knowledge and points to the promise of novice teachers. Results from the present investigation suggest that novice teachers are more effective in terms of reducing absenteeism and chronic absenteeism among their students than more experienced teachers. Although at the outset of this study we speculated that two reasons why novice teachers may be able to better address issues of absenteeism in their classrooms is that they are better able to relate with parents and they are being prepared in a context that prioritizes new accountability structures and pedagogical approaches, future work should replicate our findings with different samples and methods and test this underling theory of change.
Second, as part of the present investigation, we also considered the extent to which the associations between novice versus more experienced teachers and student absenteeism varied for different groups of children that include populations at greater risk of absenteeism, namely students in poor health, English language learners, students with disabilities, and students in poverty (e.g., Gottfried et al., 2019; Meng et al., 2012; Nauer et al., 2014). As part of this effort, we also considered whether the effects of novice teachers on student absenteeism were comparable across teachers or classrooms of different conditions. Examining these sources of heterogeneity yielded no significant differential associations. These null patterns of moderation are of note because they suggest that the benefits of novice teachers for student absenteeism and chronic absenteeism are far more comparable than different for students who experience different levels of risk and who experienced classrooms of different conditions. Although we found no evidence of moderation as a function of key child and classroom characteristics, these findings should not be taken as evidence that the effects of novice teachers are monolithic across all social groups. As such, additional work is warranted to further consider how other contextual factors and individual and familial resources may influence the effects of novice teachers.
Implications
Given these findings, there are several implications for policy, practice, and future research. For instance, from a policy perspective, these results could help inform the creation of teacher professional development that accurately capture the different dimensions of teacher effectiveness that extend beyond boosting student achievement. This is especially so given the positive results with regards to absenteeism. Thus, without paying attention to child attendance in ways that we design teacher professional development, we had a partially-obscured portrait of the ways in which novice teachers might contribute to student success and how more experienced teachers might benefit from development at an advanced career stage. Such efforts to better understand teacher effectiveness are already underway to develop and refine support systems that recognize this nuance (Marsh et al., 2017). However, this study underpins the need for additional research to inform the development of these potential professional supports that unearths what the specific mechanisms and practices are that novice teachers are engaging in that are reducing rates of absenteeism. Given our sensitivity checks, it appears that family engagement in schools is not the mechanism.
As another implication, from a practice perspective, results from this research make clear that absences can be reduced in extant school settings and, thus, opens the door to questions about the differential classroom contexts that are driving our observed results. Although due to data limitations our study could not examine the mechanisms that connect novice teachers with student absenteeism, the identification of such practices concentrated among novice teachers represent malleable targets for intervention and can be utilized by all teachers, regardless of experience, to help reduce student absenteeism. This leads to an implication with regards to the lack of significant findings for our second research question. From a research perspective, and as noted above, such consistent null findings require replication and extension before definitive conclusions are drawn.
Yet, from a policy and practice perspective, these findings suggest that although novice teachers have students with fewer absences and a lower likelihood of chronic absenteeism than their more experienced counterparts in early elementary years, this may not be sufficient to help curtail gaps in absenteeism in the early years of school among vulnerable populations (e.g., Gottfried et al., 2017; Morrissey et al., 2014; Van Eck et al., 2017). After all, we know that vulnerable populations, including children in poverty, are at a greater risk of absenteeism (Morrissey et al., 2014). Despite the lack of differential effects, there is clear evidence that absenteeism trajectories are most malleable in the early years of school and that long-term trajectories of absenteeism are established as early as kindergarten (Gottfried et al., 2019). Thus, attempts to curtail absenteeism should be targeted at the earliest years of schooling.
Limitations
Despite these contributions to the literature on absenteeism and novice teachers, the results of the present investigation should be interpreted in light of several limitations. First, although our statistical models included a wide range of child, family, teacher, and classroom covariates in addition to child fixed effects, interpretation of our findings requires caution given the potential for omitted time varying confounders; put another way, our results remain descriptive and not causal in nature. At the same time, however, our models were robust to a variety of analytic specifications, which lends greater confidence to our conclusions.
Second is in regards to our measure of student absenteeism. First, the way that NCES collected absence data was through teacher report rather than school records. Thus, we were not able to rely on exact absence counts. Additionally, the dataset did not provide information as to why children were absent from school. Although this type of measurement is standard practice in the study of student absenteeism (e.g., Ansari & Gottfried, 2021), future studies should consider using administrative data sources where at minimum excused and unexcused designations are provided. Doing so may provide more detailed information about the frequency and reasons for absenteeism.
Third, although we used a nationally representative sample of students who were tracked across the early elementary school years, an important next step is to see whether these associations replicate among students in the oldest grades. Such an effort is important and can provide better insight about the ways in which schools should go about addressing teacher effectiveness in addressing absenteeism. Reflecting the potential differences by grade level, work by Ladd and Sorensen (2017)—one of the few known studies to consider the links between teacher experience and student absenteeism—suggests that absenteeism is lowest among middle school students who have teachers with more years of experience. As such, continued work is necessary to disentangle the conditions under which novice and more experienced teachers are more effective in addressing issues related to absenteeism—and importantly, how this might differ by student age.
Fourth, as mentioned, we did not have any information on days present, only days absent. Given high levels of school mobility in some districts, not knowing days present could mean that we are undercounting chronic absence. Namely, for a student enrolled in half the year, missing 10 days in one semester means something more dramatic than missing 10 days over the full year. Therefore, our results should be compared to data that has days present across an entire year, so that we can assess both number of absences as well as rate of absenteeism.
Lastly, due to data limitations, our study was not able to shed light on the mechanisms that link teachers’ years of experience with students’ school absences. Such efforts are necessary, however, to shed light on the processes that differentiate the experiences of students who have teachers with different levels of experience.
With these limitations and future directions in mind, the present investigation leveraged nationally representative data to report on the links between elementary school teachers’ years of teaching experience and their students’ school absences. When taken together, results of this work make clear that: (a) novice teachers have students with fewer absences and a reduced odds of chronic absenteeism in the early elementary school years relative to more experienced teachers; and (b) reductions in absenteeism were comparable for students of different backgrounds and who experience different levels of risk. Accordingly, our findings underscore the importance of considering the role of teachers, and their experience levels, in addressing students’ school absences in the earliest years of education.
Footnotes
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) received no financial support for the research, authorship, and/or publication of this article.
