Detailing New Dangers: Linking Kindergarten Chronic Absenteeism to Long-Term Declines in Executive Functioning

Abstract

Of all elementary school years, absenteeism is at its peak during kindergarten. Although much has been established about the effects of missing kindergarten school days on achievement, nothing has yet been established on absenteeism and executive function (EF) skills. Yet developing EF skills early in school is critical, and missed in-school time might have long-term implications. To explore this link, we asked whether absenteeism in kindergarten was linked to both short- and long-term EF skill development. Using nationally representative data (N = 14,370) and employing fixed-effects modeling, we found that kindergarten absenteeism was linked to lower working memory and cognitive flexibility outcomes. The patterns varied based on definition of absenteeism, though our evidence does suggest long-term declines on EF skills seen through at least third grade.

Contrary to common beliefs about absenteeism, children in the earliest years of education are missing a substantial amount of school (Gottfried & Hutt, 2019). In fact, kindergartners are missing more school than children in any other elementary grade (Balfanz & Byrnes, 2012). Although chronic absenteeism (i.e., missing anywhere between 2 and 3 weeks or 10% of the school year, depending on definition; Gottfried, 2014; Jordan & Miller, 2017) between first and sixth grade is around 5%, in kindergarten this estimate is as high as 10%–15% (Bauer et al., 2018). An additional 14% of kindergartners are just below this chronic-absence designation (Chang & Davis, 2015; Romero & Lee, 2007). Interpreting these alarming statistics together, an estimated one-fourth of our nation’s kindergartners are missing so much school that it puts them into or in proximity of categories of absenteeism that policy makers would red-flag (Jordan & Miller, 2017).

Also contrary to common belief, missing school in kindergarten comes with a swath of negative consequences. Children who are more frequently absent in kindergarten demonstrate lower academic performance, both in the current year as well as in future years (Ansari & Pianta, 2019; Chang & Romero, 2008; Gottfried, 2014). For instance, Gottfried (2014) found that being chronically absent in kindergarten was associated with lower reading and math achievement outcomes at the end of that year. In terms of future performance, additional research has found that chronically absent kindergartners are less likely to meet state proficiency standards in third grade (Ehrlich et al., 2018) and are at a greater risk of being retained in school (Connolly & Olson, 2012). Kindergartners who miss school also tend to have less optimal socioemotional development, including lower social engagement (Gottfried, 2014). But even though numerous studies have examined the links between absenteeism and children’s academic achievement, and less (but some) work has examined absenteeism and social skills, little effort has been made to examine the links between absenteeism and executive function (EF) skills. The above is a key gap in knowledge because even though schools are increasingly being perceived as primarily academic institutions, schools and classrooms also play important roles in shaping children’s nonacademic skills, such as EF (Nguyen et al., 2020; Rimm-Kaufman et al., 2002).

Early Absenteeism and Missed School Opportunities

Academics

As described by Gottfried and Kirksey (2017), one key reason why children do not succeed when absent is due to unequal opportunities to learn. As kindergarten becomes more rigorous (Bassok et al., 2016) and as children spend more time in kindergarten settings rather than at home (Cooper et al., 2010), the costs of missing school have intensified. When considering academic outcomes, greater time on academic subjects has been correlated with higher academic performance (Georges, 2009; Heatly et al., 2015). Thus, children who miss school miss more opportunities to engage with academic content and instruction. And as kindergarten has become more academic and rigorous in conjunction with a rise in school absenteeism, children perform more poorly (Gottfried, 2017). These issues are potentially exacerbated for children from low-income households, where parents might face additional barriers (or competing priorities) that prevent them from compensating for lost in-school time (Chang & Romero, 2008; Gershenson et al., 2017). The issue of missed opportunities is also exacerbated for children with major health issues, as they may have extended absences, miss a disproportionate amount of school, and experience other complicating factors that could affect academic outcomes. Even when children miss school for health reasons, they experience lower academic outcomes, further underscoring that missing school (not out of disengagement) negatively affects opportunities to learn (Gottfried, 2009).

Behavior

As mentioned above, nonacademic outcomes have been less frequently studied in the area of school absences. Yet, conceptually, a similar missed-opportunities perspective would be appropriate here as well. Early schooling has been established as a time when young children not only gather academic information in the classroom but also develop behavioral skills (Barton et al., 2014; Haycock, 1999; Nguyen et al., 2020; Rimm-Kaufman et al., 2009; Sanders & Rivers, 1996; Wenglinsky, 2002). For instance, classroom time includes group interaction, collaboration, trial and error, and experimentation, and these have impacts on development (Henry, 2003), such as by promoting peer interactions, gaining concern and respect for others, being a courteous listener, fostering intrinsic motivation, becoming a lifelong learner, being a caring citizen, adopting cooperative approaches, remaining calm, and taking ownership of problems (Ahtola et al., 2011; Caldarella et al., 2015; Henry, 2003; Kramer et al., 2010; Marlowe and Page, 1998). It has been established that the early years, such as kindergarten, are especially critical to develop these skills (Duncan et al., 2007; Posner & Rothbart, 2000). Therefore, just like with academics, missing school makes it challenging for young children to be set on a positive educational trajectory because of all of the forgone opportunities for interaction. As chronic absenteeism is linked with disengagement (Gottfried, 2014), missing copious amounts of school only worsens the opportunities to develop these behavioral skill sets.

Executive Function

Although much of the extant literature has focused on children’s academic achievement and social skills in the context of education, developmental and educational research has made great progress in understanding the antecedents and outcomes of EF, which represents domain-general skills that include three distinct components: (a) working memory, (b) cognitive flexibility, and (c) inhibitory control (Blair, 2016; Diamond et al., 2007). Given data limitations, as part of the present study we focus on the first two dimensions of EF. As brief background, however, we define each of these components below. First, working memory captures children’s ability to store and manage information during complex cognitive tasks. Second, cognitive flexibility captures a child’s ability to switch between thinking about different concepts. And, finally, inhibitory control refers to a child’s ability to inhibit their own response to distractions. These dimensions of EF skills have been found to develop through young adulthood, but they most rapidly develop in the early childhood years (Blair & Razza, 2007; Garon et al., 2008). Developing EF skills from the very beginning of children’s formal education is particularly important for policy and practice, as EF skills are not only malleable (Duncan et al., 2018; Jennings & DiPrete, 2010) but are also predictive of children’s school success (Nguyen & Duncan, 2018; Rosen et al., 2010), adult life success (Cunha & Heckman, 2008), and engagement in other healthy behaviors (Chiteji, 2010).

Despite the lack of focus on the links between absenteeism and children’s EF skills, there is ample evidence to suggest that the predictability and structure of children’s early experiences matter for their EF development (Shonkoff, 2011). Indeed, developmental and educational researchers have demonstrated the importance of children’s environments, such as the school, for their development of EF skills (e.g., Nguyen et al., 2020; Pianta et al., 2020; Weiland et al., 2013). For instance, EF skills have been found to develop in the context of high-quality and child-structured classrooms (Ansari & Purtell, 2017; Fuhs et al., 2013; Goble & Pianta, 2017; Pianta et al., 2020; Weiland et al., 2013) that facilitate high-quality interactions between children and their classmates and teachers. Such findings have also been supported by recent randomized controlled trials indicating that specific aspects of education do produce improvements in children’s EF (Duncan et al., 2018; Raver et al., 2011). Even though the extant literature in this area is still in the early stages of development, it has nonetheless suggested that educational systems have agency in being able to improve children’s EF outcomes.

Accordingly, to the extent that schools facilitate EF development, it would seem logical that absenteeism may minimize the benefits of early schooling as this (a) introduces unpredictability in children’s educational experiences and (b) affords children fewer opportunities to develop these skills in the context of high-quality environments. And yet, among all characteristics, policies, and programs that have been examined to boost children’s development of EF, very few have focused directly on the role of absenteeism. Rather, the focus of reducing absences has been mostly tied to evaluating academic achievement. Studying the nonacademic consequences of school absences can help practitioners and policy makers design educational settings to be more conducive to improving a range of important outcomes.

The Current Study

To address this void and to further document the reach of chronic absenteeism, we asked the following research questions:

1. Do chronically absent kindergartners have different EF outcomes (i.e., working memory, cognitive flexibility) at the end of kindergarten as compared with children who were not chronically absent?

2. Do the links between kindergarten absenteeism and children’s EF development persist through the early elementary school grades?

3. To what extent do the outcomes of absenteeism vary by key subgroups of children?

Given the literature documented above, we hypothesize that more absences correlate with weaker EF outcomes.

Method

Source of Data

To address our research questions, we utilized nationally representative data from the Early Childhood Longitudinal Study, Kindergarten Class of 2010–11 (ECLS-K: 2011). The ECLS-K: 2011 was created by the National Center of Education Statistics (NCES) at the US Department of Education. The NCES collected data on a sample of kindergarteners through direct assessments of these children as well as through interviews of their parents, teachers, and school administrators. The NCES utilized a three-stage stratified sampling design to ensure that the sample was nationally representative: The first sampling unit was geographic region, then public and private school, and finally children stratified by race/ethnicity as the third sampling unit.

Data were initially collected on children during fall and spring of the 2010–11 school year. All independent variables for this study were sourced from both of these kindergarten-year waves. Relevant to this study, data were also collected on the sample of children in the spring of each grade. In this study, dependent variables on EF, as described below, are sourced from the fall and spring waves of kindergarten and then the spring of each subsequent grade through third grade. To create our final analytic sample of independent and dependent variables, chained multiple imputation was employed (Royston, 2004) to replace missing values (missingness for the variables in the ECLS-K: 2011 data set ranged from 0.05% to 36.5%). We imputed 20 data sets to resemble the distribution of the observed variables. We used sample weights (W12AC0) included in data set in the imputation and analyses to follow. After imputation, our sample consisted of approximately n = 14,370 kindergartners who were then followed through elementary school. Sample sizes are rounded to the nearest tens digit, per NCES rules of using the restricted version of these data.

EF Outcomes

Although there is no gold standard for measuring EF, in kindergarten through third grade, children were administered an assessment battery of tests that have been extensively used and validated with school-age children (Beck et al., 2011; Blair & Razza, 2007; Morgan et al., 2019; Tourangeau et al., 2015; Woodcock et al., 2001; Zelazo, 2006). These assessments captured children’s working memory and cognitive flexibility. Working memory was measured with the numbers reversed (NR) subtest of the Woodcock-Johnson III (Woodcock et al., 2001). A child’s score is based on reciting numbers in the reverse order. The NR subtest has been shown to be a reliable test of working memory, with a reliability of .87 reported in a sample by Schrank (2011). For the purposes of the present study, we use the W score, which is designed specifically for correlation and regression analyses (Tourangeau et al., 2015). Cognitive flexibility was measured with the Dimensional Change Card Sort (DCCS; Zelazo, 2006), which has a high test-retest reliability, ranging from .90 to .94 (Beck et al., 2011). The DCCS score reflects a child’s performance over a series of tests associated with accuracy on card-sorting tasks (i.e., by shape, color, and border games). Note that for kindergarten and first grade, the DCCS was administered with physical cards and trays. Because these assessments would have been too easy for the study children beyond this grade level, in second and third grade, children were administered the DCCS via a tablet. Differences in assessment reflect the logical progression of children’s development of EF. Moreover, for all outcomes, a higher score represents a greater mastery of the tasks at hand, hence “higher” executive functioning.

We utilized both the numbers reversed (W score) and DCCS scores as outcomes throughout this study, and all information about these verified scales is provided in the publicly available user’s manual (Tourangeau et al., 2015). As noted above, these assessments were administered in the fall and spring of kindergarten and subsequently in the spring of first through third grade. For the purposes of this study, our outcomes are derived from the spring of each grade, though we do control for the fall of kindergarten assessment in each regression model to follow as consistent with prior research (DeCicca, 2007; Gottfried et al., 2019), and, thus, our analyses consider the extent to which absenteeism is associated with changes in executive functioning over time.

Kindergarten Chronic Absenteeism

Table 1 presents all independent measures utilized in this study—and as mentioned above, all are sourced from the fall and spring kindergarten wave assessments and interviews. To begin, our key independent variable in this study was whether a child was considered to be chronically absent in kindergarten. In the spring of kindergarten, a child’s teacher reported the number absences the child had at that point in the school year. The teacher selected from a discrete set of answer choices: 0, 1–4, 5–7, 8–10, 11–19, or 20 or more. These categories are used in other research (Chang & Romero, 2008; Morrissey et al., 2014).

Table 1.

Descriptive Statistics (N = 14,370)

	Chronically Absent		Not Chronically Absent
	Mean	SD	Mean	SD	% Missing
Child characteristics:
Male	.51	.50	.51	.50	.002
Black	.15	.36	.13	.34	.005
Hispanic	.25	.44	.23	.43	.005
Asian	.06	.25	.07	.28	.005
Other	.08	.28	.06	.23	.005
English-language learner	.15	.36	.17	.39	.117
Poor health rating	.20	.40	.13	.34	.271
Educational experiences:
Full-day kindergarten	.84	.36	.82	.38	.182
Public kindergarten	.89	.31	.86	.34	.094
Hours (per week) in center-based prekindergarten	13.27	13.28	15.93	14.18	.267
Hours (per week) in before-/after-school care during kindergarten	5.96	10.99	6.12	9.85	.268
Out-of-home care prior to prekindergarten	.41	.49	.52	.50	.268
Household characteristics:
Two-partner household	.69	.45	.77	.41	.257
Number of siblings	1.49	1.20	1.49	1.12	.263
Low-income	.64	.49	.49	.50	.285
Mother has at most high school degree	.34	.47	.32	.47	.119
Father has at most high school degree	.18	.39	.21	.40	.120
Mother employed part-time	.24	.43	.27	.44	.275
Father employed part-time	.05	.22	.05	.21	.279
Number of books at home	72.91	117.27	82.22	133.46	.275
Home activities (scale)	2.99	.48	2.93	.46	.279
School-going and routines:
Bus to school	.32	.46	.33	.46	.253
Distance	5.13	4.99	4.91	3.92	.255
Time to school	12.72	7.76	12.31	7.60	.258
Regular breakfasts at home	5.27	1.75	5.38	1.68	.269
Regular dinners at home	5.47	1.91	5.50	1.81	.269
n	1,830		12,540

To determine if a child was chronically absent, we created a binary measure indicating if a child had missed 11 or more days of school (i.e., falling into the category of 11–19 or 20 or more absences) as consistent with prior research (Allensworth & Easton, 2007; Gottfried, 2014, 2015, 2017; Gottfried et al., 2016). The reference group of not being chronically absent consists of students who fell into the 0, 1–4, 5–7, or 8–10 categories. Table 1 divides the sample into those who were and were not chronically absent in kindergarten. Approximately 12% of the sample was chronically absent. Even though chronic absenteeism is often defined as missing 10% of the entire school year, we could not use this measure for two reasons. First, days of the school year were not provided in the data set, and therefore we could not calculate the exact percentage of absenteeism. Second, the spring teacher survey was administered in March of the school year rather than in June. Therefore, we assumed that children who missed more than 10 days of school by March would be either chronically absent or just shy of being so by June.

Control Variables

To reduce the possibility of spurious associations, we control for a rich set of covariates that are informed by the prior studies on school absenteeism (Ansari & Purtell, 2018; Gershenson et al., 2017; Gottfried, 2014). These control variables are outlined in Table 1. First, we included child characteristics, namely gender, race/ethnicity, English-language learner status, and an indicator for whether a parent rated the child as having poor health. Second, we included variables about a child’s educational experiences. These included indictors for having attended full-day kindergarten and whether the kindergarten attended was in a public school. In addition, we included the number of hours (from 0 and higher) that the child spent in center-based prekindergarten and hours in before-/after-school care during the kindergarten year. Finally, we included an indicator for whether the child was in out-of-home care in the year before prekindergarten.

The next category of control variables in Table 1 pertain to a child’s household characteristics. First, to account for size, we included an indicator for whether the child lived in a two-partner household as well as number of siblings. Second, we included a measure for whether the family was in a low-income scenario, and this was proxied by whether the family received government food stamps over the past 12 months. Third, we included indicators for parental education and employment. Finally, we included the number of books at home as well as a scale (Votruba-Drzal et al., 2008) for the number of learning activities that took place at home (i.e., 10 survey questions on a 4-point Likert scale regarding how often in a typical week parents engaged their child in learning activities such as reading books). Our final set of control variables captured school-going and routines. As part of these controls, we included whether the child took the school bus to school and how far the child lived from school, in miles. We also included the number of breakfasts and dinners that the family regularly had together at home, as routines have been cited as important for reducing absenteeism in kindergarten (Gottfried, 2015, 2017).

Generally, the characteristics between kindergartners who are and are not chronically absent were similar, but there are some key differences (see Table 1). For instance, a greater percentage of chronically absent children have a poor health rating and are of lower income compared with children who were not chronically absent in kindergarten. This is logical, given the high correlation between both health and income with school attendance. In addition, attending center-based care and before-/after-school care in kindergarten is linked to fewer absences. Hence, these three key differences also motivate the factors we explored in our third research question.

Analysis Strategy

We began our analysis with the following baseline regression model:

$${\text{EF}}_{\mathit{\text{ijst}}}={\beta }_{\circ }+{\beta }_1{\text{ABS}}_{\mathit{\text{jk}}}+{\beta }_2{C}_i+{\beta }_3{K}_i+{\beta }_4{H}_i+{\beta }_5{R}_i+{\epsilon }_{\mathit{\text{ijst}}},$$ (1)

where, in this model, EF represents our two EF outcomes for child i who enrolled in kindergarten classroom j in school s in spring of year t (i.e., outcomes in spring K, 1, 2, and 3). Our independent variable of interest is ABS, which is a binary measure for whether a child was chronically absent in kindergarten year k. The regression terms C, K, H, and R represent the sets of control variables described above, namely child characteristics (C), educational experiences (K), household characteristics (H), and routines (R). Finally, the error term is clustered by kindergarten classroom. Variations on the clustering of the error (i.e., school) did not change our findings below.

Model adjustments.

A first issue is that, in the model above, we are comparing the EF of kindergartners who were and were not chronically absent across the entire data set, which represents kindergartners across the United States. That said, the experiences of children in specific schools might differ dramatically, and therefore, comparing children across schools may not provide as robust of an estimate. For an example as to why, schools in California are being held accountable through its state Every Student Succeeds Act (ESSA) plan for both reducing chronic absenteeism as well as improving nontesting outcomes of their students. In this case, the effect of chronic absenteeism in kindergarten might be underestimated on EF outcomes, particularly when comparing students in California to students in Florida, where the state has chosen not to emphasize chronic absenteeism reduction nor nonacademic outcomes in their state accountability ESSA plans (Jordan & Miller, 2017). Comparing schools to other schools does not account for this observed (or unobserved) variation. As such, we revised the above model to include school fixed effects:

$${\text{EF}}_{\mathit{\text{ijst}}}={\beta }_{\circ }+{\beta }_1{\text{ABS}}_{\mathit{\text{jk}}}+{\beta }_2{C}_i+{\beta }_3{K}_i+{\beta }_4{H}_i+{\beta }_5{R}_i+{\delta }_s+{\epsilon }_{\mathit{\text{ijst}}},$$ (2)

where δ_s represents school fixed effects—that is, a series of indicators for the school the child attended. By including these indicators, both observed and unobserved school characteristics are being held constant. Thus, we are only comparing children to others in the same school, and therefore controlling for between-school differences, such as policies, school leadership, attendance programs, and efforts to boost nonacademic outcomes.

After examining the findings from this model, we revised the specification once more. In the following equation, we replaced school fixed effects with classroom fixed effects:

$${\text{EF}}_{\mathit{\text{ijst}}}={\beta }_{\circ }+{\beta }_1{\text{ABS}}_{\mathit{\text{jk}}}+{\beta }_2{C}_i+{\beta }_3{K}_i+{\beta }_4{H}_i+{\beta }_5{R}_i+{\delta }_j+{\epsilon }_{\mathit{\text{ijst}}},$$ (3)

where δ_j represents classroom fixed effects—that is, a series of indicators for the kindergarten classroom. Analogous to the setup of a school fixed-effects model, a classroom fixed-effects model includes indicators for the kindergarten classroom in which the child was enrolled. Classroom fixed effects improve on school fixed effects because they control for between-classroom differences (both observed and unobserved) in the same school. In other words, in the classroom fixed-effects model, we are comparing the EF outcomes of children who were in the same kindergarten classroom, where some were and others were not chronically absent. To use the example of the state of Florida once again, although it may be the case that as a whole the state is not holding schools accountable for absenteeism reduction or nontesting improvements (Jordan & Miller, 2017), it might be the case that individual teachers were focusing on these two areas. If teachers differed in their efforts within the same school, then comparing children across classrooms would provide a biased estimate of kindergarten chronic absenteeism. As another example, some classrooms might have teacher aides whereas others do not, even within the same school. If having an extra teaching resource in the room can help improve both attendance and EF, then comparing children across classrooms would not be an accurate assessment. Therefore, classroom fixed effects importantly compare children with their classmates with the same kindergarten teacher, with the same teacher’s aide, and so forth. Classroom fixed effects, in other words, fix attributes of the kindergarten classroom in the regression, so what varies—most importantly—are EF outcomes and chronic absenteeism among children in the same classroom.

Additional Tests

To ensure the robustness of our findings reported above, we estimated two supplemental models. First, we reran the classroom fixed-effects models described above, but we removed the subset of children who were never absent in kindergarten. That way, the analysis includes only those who were absent at some point in kindergarten, with some being chronically absent. Second, even though the focus of this study as well as current policy rhetoric pertains to chronic absenteeism, we also chose to explore days absent. Therefore, we reran all models with the outcome as the categories of days absent described above. For this variable, students who had 0 or 20 absences took on that value; those who had 1–4, 5–7, 8–10, or 11–19 took on the middle point (see also Gottfried, 2014). And even though this indicator of days absent was collected on an ordinal scale with six discrete options, prior simulation studies have shown these types of scales can be treated as continuous as long as they have four or more categories (Bentler & Chou, 1987). Accordingly, in our analyses, we treat days absent continuously.

Results

Baseline Results

Table 2 presents the findings based on employing the baseline model in Equation (1) above. Each column represents a unique regression, where the outcome is designated at the top of each column. Coefficients are presented with clustered standard errors in parentheses. The first grouping pertains to working memory and the second grouping is for cognitive flexibility. Recall that all models include the fall kindergarten measure of the outcome as a control variable in addition to other characteristics listed in Table 1.

Table 2.

Baseline Models

	W Ability Score				DCCS
	Spring K	Spring 1	Spring 2	Spring 3	Spring K	Spring 1	Spring 2	Spring 3
Chronically absent	−2.03⋆⋆ (.70)	−1.92⋆ (.82)	−2.47⋆⋆ (.90)	−1.98⋆ (.88)	−.12 (.08)	.04 (.07)	−.05 (.05)	−.08 (.06)
Child characteristics:
Male	−1.47⋆⋆ (.46)	−1.76⋆⋆⋆ (.48)	−.79 (.49)	−1.90⋆⋆ (.59)	−.20⋆⋆ (.05)	−.11⋆ (.05)	−.16⋆⋆⋆ (.03)	−.05 (.04)
Black	−4.72⋆⋆⋆ (.95)	−4.76⋆⋆⋆ (.97)	−4.35⋆⋆⋆ (.98)	−3.58⋆⋆ (1.14)	−.63⋆⋆⋆ (.11)	−.53⋆⋆⋆ (.10)	−.41⋆⋆⋆ (.07)	−.25⋆⋆⋆ (.07)
Hispanic	−3.33⋆⋆⋆ (.73)	−1.40 (.75)	1.03 (.81)	.87 (.92)	−.22⋆⋆ (.08)	−.18⋆ (.08)	.01 (.05)	.00 (.06)
Asian	2.27 (1.21)	6.07⋆⋆⋆ (1.17)	4.64⋆⋆⋆ (1.10)	5.96⋆⋆⋆ (1.14)	.03 (.11)	.01 (.11)	.10 (.10)	.26⋆⋆ (.09)
Other	.26 (1.05)	.23 (1.00)	−.05 (1.13)	1.01 (1.25)	−.01 (.11)	−.03 (.10)	.10 (.07)	.04 (.08)
English-language learner	−2.36⋆⋆ (.90)	−2.56⋆⋆ (.92)	.27 (.92)	1.65 (.93)	−.27⋆⋆ (.09)	−.02 (.09)	.03 (.06)	.08 (.07)
Poor health rating	−3.54⋆⋆⋆ (.80)	−3.00⋆⋆⋆ (.74)	−2.67⋆⋆ (.88)	−1.78⋆ (.90)	−.27⋆⋆ (.09)	−.32⋆⋆⋆ (.08)	−.22⋆⋆⋆ (.06)	−.36⋆⋆⋆ (.07)
Fall kindergarten measure of outcome	−.53⋆⋆⋆ (.01)	.33⋆⋆⋆ (.01)	.30⋆⋆⋆ (.02)	.30⋆⋆⋆ (.02)	.23⋆⋆⋆ (.01)	.16⋆⋆⋆ (.01)	.11⋆⋆⋆ (.01)	.09⋆⋆⋆ (.01)
Educational experiences:
Full-day kindergarten	2.86⋆⋆⋆ (.72)	1.69⋆ (.73)	1.19 (.92)	1.81⋆ (.89)	.15 (.09)	−.15⋆ (.07)	−.09 (.05)	−.04 (.06)
Public kindergarten	−1.03 (.83)	−1.10 (.80)	−.48 (.83)	.23 (1.18)	−.08 (.08)	.02 (.09)	−.09 (.06)	−.09 (.07)
Hours of center-based prekindergarten care	−.01 (.02)	.01 (.02)	.01 (.03)	.01 (.03)	.00 (.00)	.00⋆ (.00)	.00⋆ (.00)	−.00 (.00)
Hours of before-/after-school care during kindergarten	−.01 (.03)	.03 (.03)	−.01 (.04)	−.01 (.03)	.00 (.00)	−.01⋆ (.00)	.00 (.00)	.00 (.00)
Out-of-home care prior to prekindergarten	.02 (.56)	.34 (.51)	.28 (.58)	−.89 (.60)	.03 (.07)	.08 (.06)	.03 (.04)	.00 (.04)
Household characteristics:
Two-partner household	1.11 (1.03)	.38 (1.04)	1.57 (1.00)	.72 (1.06)	.16 (.11)	.04 (.12)	.08 (.07)	.03 (.08)
Number of siblings	−43 (.24)	−.32 (.23)	.41 (.28)	.02 (.31)	−.00 (.02)	.00 (.03)	.01 (.02)	.01 (.02)
Low-income	−2.48⋆⋆⋆ (.68)	−2.06⋆⋆ (.66)	−1.29 (.67)	−1.44 (.74)	−.16⋆ (.07)	−.16⋆ (.07)	−.06 (.05)	−.05 (.05)
Mother has at most high school degree	.27 (.51)	.87 (.59)	−.12 (.60)	−.18 (.68)	−.12 (.06)	.05 (.05)	.08⋆ (.04)	−.02 (.05)
Father has at most high school degree	−.24 (.63)	.38 (.59)	−.98 (.69)	−1.28 (.73)	.05 (.07)	−.04 (.06)	−.03 (.04)	.01 (.05)
Mother employment part-time	.56 (.63)	2.14⋆⋆⋆ (.61)	1.37 (.78)	1.63⋆ (.73)	.05 (.07)	.29⋆⋆⋆ (.07)	.03 (.06)	.08 (.06)
Father employed part-time	.11 (1.44)	.39 (1.70)	−2.57 (2.25)	−1.05 (1.80)	−.05 (.15)	.13 (.16)	.12 (.11)	.02 (.13)
Number of books at home	.00 (.00)	−.00 (.00)	−.00 (.00)	.00 (.00)	.00⋆⋆⋆ (.00)	.00 (.00)	.00⋆⋆ (.00)	.00⋆ (.00)
Home activities	1.57⋆⋆ (.58)	.81 (.61)	1.36⋆ (.59)	.28 (.78)	.07 (.07)	.06 (.05)	.08 (.05)	.08 (.06)
School-going and routines:
Bus	−1.26 (.70)	.18 (.69)	−1.39 (.71)	−1.01 (.71)	−.14 (.07)	.04 (.07)	−.09 (.05)	−.18⋆ (.07)
Distance	.11 (.06)	.15⋆ (.08)	.09 (.08)	.07 (.09)	−.01 (.01)	.01 (.01)	.01 (.00)	.01 (.01)
Time to school	−.02 (.04)	−.11⋆⋆ (.04)	−.06 (.05)	−.05 (.04)	.00 (.01)	−.01 (.00)	.00 (.00)	.00 (.00)
Regular breakfasts at home	.14 (.17)	.23 (.18)	.12 (.19)	−.10 (.18)	.00 (.02)	.01 (.02)	.01 (.01)	.02 (.02)
Regular dinners at home	−.15 (.15)	−.24 (.18)	−.19 (.16)	−.18 (.15)	−.01 (.02)	−.01 (.02)	−.01 (.01)	−.02 (.01)
n	14,370	14,370	14,370	14,370	14,370	14,370	14,370	14,370

Note.—DCCS = Dimensional Change Card Sort; standard errors adjusted for clustering in parentheses.

^⋆p < .05.

^⋆⋆p < .01.

^⋆⋆⋆p < .001.

The key variable of interest is found in the first row of Table 2—an indicator variable for whether the child was chronically absent in kindergarten. Although being chronically absent in kindergarten was not associated with students’ cognitive flexibility, chronic absentees did demonstrate less optimal working memory across all grades. Although the unstandardized coefficients are presented in the table, these translate to standardized betas of approximately −.08 (in grades K, 1, and 3) to −.12 (in grade 2). To further put these findings in perspective, we convert these associations to months of development by benchmarking these findings as a function of how much children improved in working memory across time (differences in children’s performance across time divided by time elapsed between assessments). In doing so, we find that chronically absent children demonstrated roughly 1–1.5 fewer months’ gains in working memory than nonchronically absent children between kindergarten and third grade. There is thus a great deal of consistency in these models, suggesting there is no evidence of fade-out of early chronic absenteeism.

Although the focus of this study was on the links between chronic absenteeism and children’s EF, we briefly describe the associations between our control variables and EF. Throughout the table, there were consistent results that EF outcomes were differentiated by child characteristics, including boys, Black and Hispanic children, English-language learners, and those who are less healthy having lower EF scores. Asian children, however, had higher working memory scores across grades. Children in two-partner households tended to have higher EF scores, and those in low-income families tended to have lower scores. There are no consistent findings for other household or school-going and routines variables.

Model Adjustments

In Table 3, the findings are presented for our school and classroom fixed-effects models, as described by Equations (2) and (3) above. Each cell represents the coefficient and standard error from a unique regression based on the outcome and specification listed. For simplicity, only the results for chronic absenteeism are shown, though all variables from Table 2 are included.

Table 3.

Fixed-Effects Models

	Spring K	Spring 1	Spring 2	Spring 3
Outcome: W ability score:
School fixed effects	−2.36⋆⋆ (.76)	−1.91⋆ (.89)	−2.45⋆ (.99)	−2.00⋆ (.94)
Classroom fixed effects	−2.08⋆ (.93)	−1.96⋆ (.94)	−1.83⋆ (1.06)	−1.42 (1.09)
Outcome: DCCS:
School fixed effects	−.16 (.08)	−.01 (.07)	−.08 (.06)	−.11 (.06)
Classroom fixed effects	−.15 (.10)	.03 (.09)	−.03 (.07)	−.09 (.07)

Note.—DCCS = Dimensional Change Card Sort; standard errors adjusted for clustering in parentheses.

^⋆p < .05.

^⋆⋆p < .01.

The school fixed-effects models are presented in the first row of Table 3. The interpretation of these coefficients is consistent with our baseline specification. Hence, there do not appear to be any between-school biases that affected our baseline estimates. When looking at the classroom fixed-effects models, however, the estimates and patterns have slightly tempered. In addition, the coefficient in spring of third grade is no longer statistically significant. Thus, there was some degree of overestimation in the estimates when relying on the baseline and school fixed-effects model. That said, the results for working memory still suggest that children who were chronically absent in kindergarten have lower working memory over time, and this is true through at least second grade.

Additional Tests

As discussed above, the first additional test removed the subset of children who were never absent. The results in Panel A of Table 4 are consistent with the classroom fixed effects coefficients reported in Table 3. Hence, including those children who were never absent in kindergarten did not skew the findings from the classroom fixed-effects models in Table 3. In Panel B of Table 4, we replaced the binary indicator for being chronically absent for the number of days absent per child in kindergarten. Using the classroom fixed-effects model, the findings suggest that missing more days of school was associated with lower working memory as well as cognitive flexibility—through third grade for both outcomes. For working memory, the coefficients across grades translate to standardized coefficients of −.07 to −.11. For cognitive flexibility, the coefficients across grades translate to standardized betas of about −.04 in kindergarten and first grade, −.14 in second grade, and −.11 in third grade.

Table 4.

Tests of Robustness

	Spring K	Spring 1	Spring 2	Spring 3
	A. Removing Students with Zero Absences
Outcome: W ability score:
Classroom fixed effects	−2.03⋆ (.95)	−1.98⋆ (.96)	−1.84⋆ (1.09)	−1.33 (1.11)
Outcome: DCCS:
Classroom fixed effects	−.18+ (.11)	.00 (.09)	−.03 (.06)	−.06 (.08)
	B. Days Absent
Outcome: W ability score:
Classroom fixed effects	−.45⋆⋆ (.14)	−.45⋆⋆ (.14)	−.53⋆⋆⋆ (.15)	−.38⋆ (.16)
Outcome: DCCS:
Classroom fixed effects	−.02⋆ (.01)	−.02⋆ (.01)	−.04⋆⋆⋆ (.01)	−.03⋆⋆ (.01)

Note.—DCCS = Dimensional Change Card Sort; standard errors adjusted for clustering in parentheses.

⁺p < .10.

^⋆p < .05.

^⋆⋆p < .01.

^⋆⋆⋆p < .001.

Group Differences

In this final section, we considered whether there were key group differences in working memory and cognitive flexibility for children who were chronically absent in kindergarten. To do so, we utilized the classroom fixed-effects model and interacted the chronic absence indicator with one of three key characteristics.

Table 5 presents the findings for these partially interacted models. Each grouping represents the coefficient and standard error from a regression model where the chronically absent indicator was interacted with the key characteristic presented in the table. Any statistically significant coefficient would indicate a unique relation for less-healthy children, lower-income children, or children who attended prekindergarten. Nothing emerged as consistently significant for an interaction. Though not shown, we also interacted chronic absenteeism with fall ability score, but the results were also not statistically significant. Thus, all children did less well when they were chronically absent. Finally, the bottom portion of the table repeats the analysis using days absent rather than chronic absence, and similar patterns emerge.

Table 5.

Differences by Group

	Spring K	Spring 1	Spring 2	Spring 3
	A. Chronic Absenteeism
Outcome: W ability score:
Poor health rating * chronic absence	−4.19 (2.30)	−1.67 (2.56)	−2.80 (2.90)	−.80 (2.89)
Low-income * chronic absence	−1.44 (1.81)	−1.08 (1.92)	.80 (2.10)	−.03 (2.29)
Center pre-K * chronic absence	−.82 (1.73)	1.34 (2.06)	−.36 (2.22)	.29 (2.00)
Outcome: DCCS:
Poor health rating * chronic absence	.20 (.24)	−.05 (.24)	−.04 (.17)	.02 (.18)
Low-income * chronic absence	−.15 (.19)	.05 (.19)	−.08 (.12)	−.12 (.14)
Center pre-K * chronic absence	−.03 (.19)	.07 (.18)	−.06 (.12)	−.18 (.14)
	B. Days Absent
Outcome: W ability score:
Poor health rating * number of days absent	−.35⋆ (.16)	−.19 (.17)	−.29 (.20)	−.18 (.20)
Low-income * number of days absent	−.15 (.14)	−.20 (.15)	.00 (.14)	−.11 (.15)
Center pre-K * number of days absent	.01 (.01)	.00 (.01)	−.01⋆ (.01)	−.01⋆ (.01)
Outcome: DCCS:
Poor health rating * number of days absent	.00 (.02)	−.01 (.02)	−.01 (.01)	−.01 (.01)
Low-income * number of days absent	−.01 (.01)	.00 (.01)	−.01 (.01)	−.02 (.01)
Center pre-K * number of days absent	.01 (.01)	.00 (.01)	−.01 (.01)	−.01⋆ (.01)

Note.—DCCS = Dimensional Change Card Sort; standard errors adjusted for clustering in parentheses.

^⋆p < .05.

Discussion

Policy, research, and practice have recently sought to minimize the individual and societal costs of early absenteeism on children’s academic and sociobehavioral outcomes (e.g., Ansari & Purtell, 2018; Chang & Romero, 2008; Ehrlich et al., 2018; Gottfried, 2014; Jordan & Miller, 2017). But among those studies that have examined the short- and/or long-term ramifications of missing school, little attention has been paid to whether kindergarten absenteeism is also linked to the development of children’s EF skills. From a research perspective, this is a key gap in knowledge on EF as well as on absenteeism, given how damaging chronic absence has been shown to be on early outcomes and how critical EF skills are for academic and human development, as described above. By not considering EF skills as an outcome of absenteeism, our vantage point on the negative links between absenteeism and child outcomes has been obscured. This certainly affects our policy and practice perspectives. A majority of states are holding schools accountable for chronic absenteeism reduction and/or improvements in nontesting outcomes (Jordan & Miller, 2017). An unrefined portrait of risk from missing school might inhibit decisions to be made around reducing absenteeism as well as improving children’s growth.

Accordingly, the present study sought to address this void by first inquiring whether chronically absent kindergartners have different EF outcomes at the end of kindergarten as compared with children who were not chronically absent. Second, we looked at whether the links between kindergarten absenteeism and children’s EF development persist through the early elementary school grades. And, finally, we considered the extent to which the outcomes of absenteeism vary by key subgroups of children—characteristics that have been previously established as important differentiators in the absence literature (e.g., Gottfried, 2015; Nauer et al., 2014). In addressing these research questions, several noteworthy findings emerged that we discuss in more detail below.

To begin, our work extends what is known about the outcomes of absenteeism. We have demonstrated that children who are absent in kindergarten have lower optimal working memory and lower cognitive flexibility (when considering “days absent” instead of “chronically absent”) by the end of the kindergarten school year. Accordingly, these results affirm that even though policy and practice might embrace the academization of kindergarten, schools are more than academic institutions. An un-weighted focus on children’s achievement underestimates the breadth of missed school opportunities spurred by absenteeism. Prior research on absenteeism has highlighted that the predominant mechanism by which we have thought about the opportunity costs of absenteeism is through school days as missed opportunities to learn, hence reflected in declines in achievement (Gottfried & Kirksey, 2017). The findings in the present study, however, also suggest that these opportunity costs surface not simply in missed opportunities to learn but also missed opportunities to develop. Such findings align with recent policy efforts that consider both academic and socioemotional/EF improvements in school performance (Melnick et al., 2017). If we expect schools to foster academic and nonacademic development, then it would be logical that if children miss school, they miss other opportunities to grow.

Unlike most of the extant literature that has focused on end-of-year outcomes of absenteeism, the present study also considered whether there are persisting effects of early absenteeism. We find that children who were chronically absent in kindergarten continued to do less well in areas of working memory (but not cognitive flexibility) as compared with those who were not chronically absent. These findings run contrary to what some parents believe about early absenteeism. Therefore, an expanded perspective on missed opportunities is particularly noteworthy given the role of fostering EF for long-term life success and well-being (e.g., Heckman, 2008; McClelland et al., 2013). If taken at face value, and given the long-term implications of children’s EF skills, what these results suggest is that missing kindergarten has long-standing ramifications—in this study shown through the first portion of elementary school and perhaps even longer. Though we cannot pinpoint the precise reasons why chronic absenteeism was associated with less optimal EF skills in both the short and long term, we can speculate on the potential mechanisms given what we know about the malleability and development of EF. More specifically, it is likely that chronic absenteeism interferes with children’s development of EF skills because it introduces unpredictability in children’s educational experiences (Shonkoff, 2011) and affords children fewer opportunities to develop these skills in high-quality environments (Pianta et al., 2020; Weiland et al., 2013).

Finally, our test of robustness with regards to days absent highlights a critical issue when considering research on absenteeism. When we replaced chronic absenteeism with days absent, there were links to both working memory and cognitive flexibility. In fact, the coefficients were statistically significant from kindergarten through third grade for both outcomes, even in the most restrictive of models. This highlights an important issue for both measurement and policy with regards to absenteeism. As discussed by Gottfried and Hutt (2019), a focus on chronic absenteeism cutoffs is not necessarily motivated by choices in data, though it is used so widely that the concept of chronic absenteeism has truly become the currency in the field. However, even though the notion of “chronic” absenteeism does provide a sense of urgency as well as a digestible construct to convey to the public, there is nothing empirically unique about a cutoff. This has been exemplified in this study, as well as with others (Gershenson et al., 2019), who find that every day of missed school matters, starting from the first one. Therefore, a focus exclusively on chronic absence may be a bit rigid and perhaps obscures the effects of missing school, which is exemplified in this current study by the fact that “days absent” coefficients had comparable (or larger) effect sizes to the ones for chronic absenteeism. Although measurement is certainly not the focus of this study, a secondary implication to this work is that though chronic absenteeism is a warning sign, we cannot ignore the absence behaviors of children who fall below a threshold.

Limitations and Future Research

There are some limitations in this study that can be used to guide future research. First, even though this study relied on direct assessments from kindergarten through third grade, we could not determine the full extent of the long-term link between absenteeism and children’s EF and educational success. Future research might rely on a data set that would allow for such long-term comparisons, possibly into high school and adulthood. Perhaps these potential data sets might not be nationally representative or as large scale as what was used in this study. Instead a much smaller but more longitudinal sample might suffice to test these relations. As an alternative example, as states begin to track and measure student skills like socioemotional learning and EF, perhaps a longitudinal study of the link between absenteeism and EF will be possible in a decade or so after these data were collected.

Second, although we do find a link between absenteeism and EF, it is not possible to determine the precise mechanism as to why. Accordingly, future studies should more carefully consider the reasons why absentees do less well in school, especially in the areas of EF. This work might take the form of qualitative research, where children who are more frequently absent are contrasted to those who are less frequently absent. Understanding absenteeism from children’s perspectives might enable for a richer discussion of what is happening when children miss school.

As a final limitation, it was not possible to address teacher, administrator, or parental response to children missing school. In other words, the ECLS-K data do not allow for an examination of how schools and families might compensate and improve EF for those children with missed in-school time. This calls for two additional studies. First, additional quantitative work could examine school or parent responses through a survey—trying to understand how, for example, teachers respond to absent children when they return to school. Second, qualitative research could be used to try to understand from the adult perspective what is being done—and what needs to be done—to minimize missed opportunities.

With these limitations and future directions in mind, the present investigation expands our understanding regarding the reach of kindergarten absenteeism. Although we know that children who are more frequently absent from school struggle in areas of academic achievement (Ansari & Pianta, 2019; Chang & Romero, 2008; Gottfried, 2014), what these results make clear is that they also do less well in core areas of EF and that these deficits persist through the early elementary school grades. Accordingly, kindergarten absenteeism is an early warning sign as it has negative and persisting ramifications for children’s development.