“Academic Redshirting” in Kindergarten

Abstract

We use two nationally representative data sets to estimate the prevalence of kindergarten “redshirting”—the decision to delay a child’s school entry. We find that between 4% and 5.5% of children delay kindergarten, a lower number than typically reported in popular and academic accounts. Male, White, and high-SES children are most likely to delay kindergarten, and schools serving larger proportions of White and high-income children have far higher rates of delayed entry. We find no evidence that children with lower cognitive or social abilities at age 4 are more likely to redshirt, suggesting parents’ decisions to delay entry may be driven by concerns about children’s relative position within a kindergarten cohort. Implications for policy are discussed.

Keywords

school readiness kindergarten academic redshirting

According to scholarly and popular reports, “redshirting” in kindergarten—the practice of delaying a child’s entry into kindergarten for a year—is increasingly common (Bazelon, 2008; Graue & DiPerna, 2000; Paul, 2010; Weil, 2007). In the past, nearly all children entered kindergarten at age 5, but a recent New York Times article noted that in 2008, 17% of kindergarten entrants were at least 6 years old (Paul, 2010). The shift toward later school entry is partially explained by changes in school-entry laws, but may also be because of an increase in the prevalence of redshirting (Deming & Dynarski, 2008). Some accounts suggest parents increasingly delay kindergarten entry because they believe their child is not developmentally ready for the increasing demands of school (Cosden, Zimmer, & Tuss, 1993; Graue, Kroeger, & Brown, 2002; Noel & Newman, 2003). Others argue that redshirting reflects parents’ strategic desire to ensure their child enters kindergarten older, taller, and with higher levels of social and cognitive skills than the child’s schoolmates (Frey, 2005; Graue et al., 2002; Matlack, 2011).

Our article attempts to fill a gap in the growing redshirting literature by carefully describing the prevalence and correlates of delayed school entry. Existing research and media reports differ widely in their estimates of the prevalence of redshirting. We are able to provide more credible estimates (and from more recent cohorts) than prior research. Understanding these patterns has important implications if decisions by some parents to delay their children’s school entry affect other children. For instance, high rates of redshirting may influence a kindergarten cohort’s age composition, which in turn might shape learning via altered curricula, instruction, or peer relations. The potential for such externalities makes an examination of individual redshirting behavior a matter of public policy interest.

We find that between 4% and 5.5% of children delay kindergarten, a number lower than those typically reported. Male, White, and high-SES children are most likely to delay kindergarten. Furthermore, at the local level we find high variation in redshirting practices, with schools serving more high-income and White children displaying higher rates of redshirting.

Background

How Prevalent Is Redshirting?

In the early half of the 20th century, as kindergarten became institutionalized as part of the public school system, the percentage of 5-year-olds enrolled rose from near 0% to 60% (Cuban, 1992). More recently, however, the trend has reversed. The percentage of 6-year-olds enrolled in first grade or above has gradually dropped from 96% in 1968 to 83% in 2010 (Deming & Dynarski, 2008; Paul, 2010). Between 1975 and 2000, 22 states increased the minimum entry age for kindergarten (Stipek, 2002). However, policy shifts in entry laws only partially explain why children are entering kindergarten at a later age. Changes in the prevalence of kindergarten “redshirting”—delaying a child’s kindergarten entry for a year for developmental or social reasons—may also play a part.

In recent years, several studies have provided estimates of redshirting based on nationally representative data. Using the Child Health Supplement to the 1988 National Health Interview Survey, Byrd, Weitzman, and Auinger (1997) find that in a national sample of children ages 7 to 17 born in the late 1970s, approximately 12% are “old-for-grade” and have never repeated grades (their proxy for redshirting). Using parent-reported data from the National Household Education Survey of 1993 and 1995, Zill, Loomis, and West (1997) report that approximately 9% of first and second graders born in the late 1980s delayed kindergarten entry. Lincove and Painter (2006), who use retrospective data from the National Education Longitudinal Study of 1988 (a cohort of students born in the mid-1970s), also report redshirting rates around 9% but base these estimates only on children with summer birthdays (for whom redshirting is likely most common). Finally, estimates from the kindergarten cohort of the Early Childhood Longitudinal Study (ECLS-K), which tracks a nationally representative sample of children who entered kindergarten in 1998 (born in the early 1990s), are lower: 5% if calculations are based on children’s age at kindergarten entry and 7% if parents’ reports are used (Datar, 2006b).

These studies, which place the national redshirting rate between 5% and 12%, rely on data from cohorts born in the mid-1970s through the early 1990s, with the lowest estimated rates for the more recent birth cohorts. Kindergarten entry behaviors may have shifted in recent years because of changes in kindergarten cutoff dates and kindergarten entry exams, increased understanding about child development, large-scale expansions in preschool opportunities, and increased school-level accountability (Barnett, Epstein, Friedman, Sansanelli, & Hustedt, 2009; Jacob, 2005; Shonkoff & Phillips, 2000; Stipek, 2002). In addition, these studies rely on retrospective parental reports or fairly crude proxies to identify redshirting, and most do not account for differences across states in school-entry laws (Datar, 2006b, is an exception). To date, none of the nationally representative studies have used prospective data that include detailed information about children’s abilities and circumstances in the years prior to their school-entry decision. The current study leverages prospective, nationally representative data to provide more recent and reliable estimates of redshirting.

Who Redshirts?

The popular press emphasizes that White children, boys, and children from families with more resources are overrepresented among redshirters, a claim confirmed by earlier research (Datar, 2006a; Dobkin & Ferreira, 2010; Zill et al., 1997). Data from the National Household Education Survey of 2007, a survey of parents with children ages 3 to 5 not yet enrolled in school, indicate that 3% of parents with a high school diploma or less planned to delay kindergarten entry, compared to 12% of parents with a bachelor’s degree, and that White parents were more than twice as likely to plan to redshirt relative to Black and Hispanic parents (O’Donnell & Mulligan, 2008).

No study we are aware of has examined whether kindergarten entry decisions relate to measures of children’s cognitive, social, and physical development before school entry. If the children who redshirt are those with the lowest academic or social skills at the time when they are eligible to enroll, their delayed entry will tend to compress the skill distribution of a kindergarten cohort. Conversely, if those who enter late are disproportionately from more advantaged backgrounds and have higher social and cognitive skills, their decisions may widen the skill distribution of a kindergarten cohort. In the current study we examine whether redshirting children appear “less ready” on a variety of parent, caregiver, and direct measures of development relative to similar-aged children who start kindergarten when eligible. We also explore whether racial, gender, and socioeconomic differences in redshirting behavior can be partly explained by differences in children’s development and school readiness.

Finally, we provide a simple descriptive analysis of the extent to which redshirting practices differ across schools nationwide. When parents make kindergarten entry decisions, they may be influenced by their neighbors and friends, preschool teachers, kindergarten teachers, and others. Norms around redshirting likely differ meaningfully across communities. Indeed, in their investigation of delayed kindergarten entry across Wisconsin school districts, Graue and DiPerna (2000) found rates varying from 3% to 94%. We therefore supplement our national description of redshirting with a school-level analysis, exploring how much redshirting varies and whether schools’ demographic composition explains this variation.

Does Redshirting Affect the Composition of Kindergartener Cohorts?

Although individual parents make kindergarten entry decisions based on the costs and benefits they perceive for their own child and family, the aggregated decisions of parents could have ramifications for the learning experiences of all children over and above any individual impacts. One mechanism by which redshirting could affect students’ experiences in kindergarten is by influencing the range of children’s cognitive, social, and physical development within kindergarten cohorts.¹ In the current article we describe how redshirting relates to the within-school distribution of kindergarten students’ ages as well as cognitive skills at school entry.

To summarize, the current article adds to the previous literature in several ways. First, we provide more reliable and current information about the extent of redshirting nationwide and the characteristics of children who delay school entry. Second, we examine whether the decision to delay kindergarten is related to children’s demographic characteristics, to measures of their prekindergarten cognitive and social development, and to school contexts. Finally, we examine the extent to which redshirting affects the composition of kindergarten cohorts within schools.

Method

Data

This article utilizes data from two cohorts of the ECLS. The ECLS Birth Cohort (ECLS-B) is the first study to track a large, nationally representative sample from birth to school entry. The base sample includes approximately 10,700 infants born in 2001. Data were collected at five time points, and each wave includes direct child assessments as well as parent interviews. Wave 4 was conducted in the fall of 2006 when the majority of the sample began kindergarten. Approximately a quarter of the sample were not yet eligible for kindergarten in their state or were age eligible but did not enroll. The fifth round of data collection tracked the subsample of children who entered kindergarten in the fall of 2007. We therefore observe each child both in the year he or she turns 5 (2006) and in the year he or she enters kindergarten (2006 or 2007), regardless of whether these occur in the same or different years.

The data provide information about each child’s month of birth as well as his or her state of residence at age 5. We use these two pieces of information, combined with data on state kindergarten cutoff dates, to determine whether the child was age eligible for kindergarten in fall 2006 and the distance (in months) between his or her birthday and the state cutoff. Data on state kindergarten cutoff dates, collected by the Education Commission of the States, are presented in Table 1. Entry cutoff dates vary from as early as July 1 (Indiana) to January 1 (Connecticut), with a modal entry cutoff date of September 1.²

Table 1

State Cutoffs for Kindergarten Entry, 2005

Cutoff	State
July 1	IN
August 1	HI
August 15	AK
August 31	DE, KS, WA
September 1	AL, AZ, FL, GA, ID, IL, MD, MN, MS, ND, NM, OK, OR, RI, SC, SD, TX, WI, WV
September 2	UT
September 10	MT
September 15	AR, IA, WY
September 30	LA, NV, TN, VA
October 1	KY
October 15	ME, NE
October 16	NC
December 1	MI
December 2	CA
December 31	DC
January 1	CT
No state cutoff	CO, MA, MO, NH, NJ, NY, OH, PA, VT

Source: Education Commission of the States, http://www.ecs.org/clearinghouse/58/28/5828.pdf.

We use data on gender, race and family socioeconomic status to examine whether the rate of compliance with kindergarten entry laws varies across groups. In addition, the data include several measures of children’s “school readiness,” including measures of cognitive and social development. The combination of measures available in the ECLS-B allows us to assess the extent to which the timing of kindergarten entry is associated with children’s cognitive, social, and physical development. The direct cognitive assessments provide measures of children’s preliteracy and mathematics ability when the children are approximately 48 months. These assessments measure skills including phonological awareness, letter recognition, print conventions, number sense, counting, and pattern understanding.

In addition to direct assessments, the ECLS-B surveyed each child’s primary care provider (preschool teacher, babysitter, etc.) and included a series of questions about the child’s behavior, social skills, and maturity. The items, which are modified from the Preschool and Kindergarten Behavioral Scales, Second Edition, asked caregivers whether the child makes friends easily, shares toys, works independently, and acts impulsively. We use factor analysis to construct a single measure from the 20 teacher-reported measures and standardize it with a mean of 0 and a standard deviation of 1. Higher values indicate better adjustment and social skills.³ Parents also answered similar questions about their child’s social skills. We constructed an analogous factor and find the correlation with the provider-reported assessments is modest (.35).

Finally, parents were asked about their child’s proficiency on a set of “basic” skills, including letter and color recognition as well as counting. We constructed a variable that ranges from 0 to 3 depending on the number of the basic skills the child could demonstrate across the following three measures: (a) child knows (at least) some of the letters of the alphabet, (b) child can identify the colors red, yellow, blue, and green by name, and (c) child can count up to 10 or higher. Just more than 70% of the children in our sample met all three of these basic skills, and an additional 18% met two of three.

For the purpose of this study, the key strength of the ECLS-B is that it provides prospective information about an age cohort of children prior to kindergarten entry and includes several direct measures of child development. It allows us to more accurately measure the correlates of delayed school entry than has previously been possible. A limitation of the data set, however, is that it does not allow us to observe groups of kindergarteners nested within the same school or neighborhood. Rates of redshirting likely differ substantially across schools and communities. To assess this, and to explore whether redshirting affects the composition of incoming kindergartener’s cohorts within their school, we use the ECLS-K, which includes data on a nationally representative sample of kindergarteners in fall 1998.

Our ECLS-K sample includes all first-time public school kindergarteners with a completed parent survey in fall 1998, a sample of approximately 12,000 children. We define each child as a redshirter if he or she entered kindergarten for the first time a year or more after he or she was first eligible, as calculated based on his or her date of birth and his or her state’s kindergarten entry cutoff in 1998 (Datar, 2006a).⁴ We aggregate the individual redshirting variable for all children attending kindergarten in the same school to construct a school-level measure, and we use the same approach to construct measures of each school’s racial and socioeconomic composition. On average, we observe 18 children per school. We explore whether schools’ redshirting rate is related to the age spread of kindergarteners in a school, defined as the difference between the 10th and 90th percentiles of the school’s age distribution. We also explore whether redshirting rates are related to the spread of school-entry academic abilities, as measured by the within-school standard deviation of literacy and mathematics test scores on the ECLS-K tests administered in fall of kindergarten.

Finally, as a check on our estimate of redshirting prevalence, we also use data from the October Supplement of the Current Population Survey (CPS). The CPS surveys a nationally representative sample of households annually; in October, parents are surveyed regarding the school enrollment status of their children (U.S. Department of Commerce, 2006). Although the CPS does not include children’s birth dates, we can estimate the proportion of redshirters by examining the proportion of 6-year-olds enrolled in kindergarten in October 2006 and adjusting for the proportion we would expect to be enrolled if all students enrolled when eligible.

Analytic Approach

How many children redshirt?

We classify ECLS-B children as redshirters if their fifth birthday falls before their state’s cutoff date for kindergarten enrollment, but they do not enroll in kindergarten until 2007 (a year after they are eligible).⁵ We classify children as on-time entrants if they are born before their state’s cutoff and enroll in kindergarten in 2006 or if they are born after their state’s cutoff and enroll in 2007. Finally, we classify children as “greenshirters” (early entrants) if they are born after their state’s cutoff and enroll in 2006. We describe the prevalence of each of these behaviors within our nationally representative sample.

Who Redshirts?

We provide basic descriptive statistics about the rates of redshirting across socioeconomic and racial groups within the ECLS-B. We then use regression to estimate the relationship between observable child and family characteristics and kindergarten entry practices for children born in the 3 months before or after their state’s kindergarten entry cutoff. For each child we calculate the difference (in months) between the child’s birth date and the kindergarten cutoff in his or her state (we call this difference D_i).⁶ We then fit logistic regression models of the form,

η_{i} = 1 n (\frac{Pr (R e d s h i r t)}{1 - Pr (R e d s h i r t)}) = β_{0} + β_{1} D_{i} + β_{2} D_{i}^{2} + X_{i} B .

We fit one model for those who turned 5 before their state’s cutoff date in 2006; this model allows us to examine the prevalence of redshirting as a function of the distance between their birth date and the cutoff date and a set of child and family covariates (X_i). We fit a separate, identical model, predicting greenshirting, for those who turned 5 after their state’s cutoff date in 2006.

Although the ECLS-B allows us to examine child-level correlates of redshirting, we use the ECLS-K to estimate the association between school-level covariates and redshirting rates. We fit a set of random-intercept logit models predicting redshirting based on schools’ racial and socioeconomic composition. In these models, we center school characteristics around their state averages, so that the coefficients correspond to average within-state associations between school characteristics and redshirting rates.

Redshirting and the composition of kindergartener cohorts

Finally, to provide some (admittedly crude) assessment of the extent to which redshirting affects the range of developmental outcomes within schools, we use the ECLS-K data and fit school-level regression models predicting the range of school-entry cognitive scores (adjusted for each child’s month of assessment) based on the rate of redshirting within the school.

Results

The Extent of Redshirting

Table 2 breaks down our sample into four categories based on kindergarten entry behaviors. The first column shows the percentage of children who turned 5 in 2006 prior to the kindergarten cutoff in their state and who, in compliance with the law, began kindergarten in 2006. The second column shows a second group of compliers who began kindergarten in 2007 because their fifth birthday fell after the 2006 state cutoff. The final columns show two types of noncompliers. The first, shown in column 3, are the greenshirters: children born after the cutoff who entered in 2006 nonetheless. Redshirters, shown in column 4, are children who were eligible by age for kindergarten in 2006 but who delayed and did not actually enroll until 2007.

Table 2

Kindergarten Entry Decisions, by Demographic Characteristics

	“On-time” kindergarten entry based on child’s birthday		Early entrants	Late entrants
	Started kindergarten in 2006	Started kindergarten in 2007	Greenshirt	Redshirt	~N
Total	70.1	24.0	2.0	3.9	5,550
Gender
Male	68.7	24.4	1.6	5.2	2,800
Female	71.5*	23.6	2.3	2.5***	2,750
SES quintiles
First quintile (lowest)	69.5	24.7	3.6	2.3	1,100
Second	68.5	26.4	1.6**	3.4	1,050
Third	73.8*	22.0	1.8**	2.4	1,100
Fourth	70.3	23.1	1.4***	5.2***	1,100
Fifth quintile (highest)	68.4	23.8	1.4***	6.4***	1,250
Race
White	69.1	24.1	1.1	5.8	2,200
Black	67.1	27.7*	4.6***	0.6***	850
Hispanic	74.0***	22.4	1.7	2.0***	1,150
Asian	74.8	16.7*	5.8***	2.7	600
Region
South	67.4	27.8	2.3	2.5	2,450
Midwest	63.4*	26.8	2.3	7.6***	1,200
West	77.5***	16.9***	1.5	4.1**	1,750

Note. Each cell is a row percentage.

Asterisks indicate significant differences between the marked group and the top category within each group. *p < .05.

p < .01. ***p < .001.

The first row of the table shows that, on average, about 94% of children comply with kindergarten entry laws (70% who enroll on time in 2006 and 24 in 2007). In all, 2% fall in the “greenshirters” category.⁷ On average, these children who enroll “early” have kindergarten experiences that are significantly different from those of “complier” children: 14% of “greenshirters” are homeschooled for kindergarten, 18% attend private schools, and 22% attend transitional classrooms (i.e., transitional kindergarten, 2-year kindergarten, pre–first grade, etc.). The comparable figures among children who entered kindergarten the year they were eligible by age are 2%, 11%, and 2%, respectively.

We find that 4% of children redshirted in 2006, according to the ECLS-B data. This is a lower rate than those reported in other studies (Datar, 2006a; Lincove & Painter, 2006; Stipek, 2002; Zill et al., 1997). As a check on this estimate of the extent of redshirting, we estimate the percentage of redshirters in fall 1998 using the ECLS-K data and the 2006 October Supplement of the CPS. Using ECLS-K data, we estimate that 5.5% of first time kindergarteners are redshirters. The CPS data show that in October 2006, the year that the ECLS-B children turned 5, 15% of 6-year-olds were enrolled in kindergarten and an additional 1% were attending preschool or prekindergarten. Some 6-year olds in kindergarten in October 2006 turned 5 after the 2005 kindergarten cutoff date in their state, however, and so enrolled on time in kindergarten in 2006. For instance, a child born on October 1, 2000, in a state with a September 1 kindergarten cutoff was not eligible for kindergarten until fall 2006, but would have been 6 at the time of the CPS survey. Based on the cutoff dates reported in Table 1, we estimate that 8% of 6-year-olds fall into this category,⁸ such that we would expect to see 8% of 6-year-olds in kindergarten in October, based strictly on their states’ cutoff. In addition, the CPS indicates that 3% of 6-year-olds are repeating kindergarten and should not be considered redshirters, as they were 5 when they first entered kindergarten. After accounting for these kindergarten repeaters and compliers, we are left with approximately 5% of 6-year-olds who should properly be classified as redshirters. This figure is well aligned with the evidence from the ECLS-B. Thus, all three of our sources suggest that redshirting rates are between 4% and 5.5%.

Who Redshirts?

Redshirting and child demographics

The remainder of Table 2 highlights differences in school-entry patterns across groups. Redshirting is twice as likely among boys than among girls. Although 2.3% of children in the lowest SES quintile delay kindergarten entrance, the figure is 6.4% among children in the highest quintile. The differences across racial groups are particularly striking. Nearly 6% of White children are classified as redshirters. In contrast, fewer than 1% of Black children delay entry. The figures for Hispanic and Asian children are only 2% and 2.7%, respectively. The patterns for greenshirting, or early entrance, are largely the reverse. Low-SES children and Black and Asian children are more likely than high-SES, White, and Hispanic children to start kindergarten before they are legally eligible.

Figure 1 expands on these findings graphically. It shows the relationships between kindergarten enrollment patterns and children’s age at the kindergarten cutoff date, based on the logistic regression models. Nearly all children born 5 or more months before their state cutoff enter kindergarten in 2006. The likelihood of starting kindergarten in 2006 drops rapidly for children more proximate to the cutoff date and among children born in the month before the cutoff, about 80% actually enroll “on time.” Figures 2 to 4 show the same information, broken down by gender, race, and SES. The gaps in the likelihood of redshirting are substantial, particularly between low- and high-SES children.

Figure 1.

Kindergarten enrollment rate, 2006, by distance of birth date from kindergarten entry cutoff date.

Figure 2.

Kindergarten enrollment rate, 2006, by gender and distance of birth date from kindergarten entry cutoff date.

Figure 3.

Kindergarten enrollment rate, 2006, by race and distance of birth date from kindergarten entry cutoff date.

Figure 4.

Kindergarten enrollment rate, 2006, by SES and distance of birth date from kindergarten entry cutoff date.

Because the bulk of noncompliance behavior occurs in the months before and after the cutoff, we focus on children born within 3 months of their states’ cutoff for the regression analyses examining the associations between child characteristics and redshirting behavior. The first two columns of Table 3 show results from a series of linear probability models predicting kindergarten entry in 2006.⁹ These models provide information about bivariate relationships between child characteristics and kindergarten entrance behaviors. Each model is numbered, and the results can be read horizontally. For instance, the first row shows the coefficients on “female” from the two models predicting redshirting and greenshirting. Among children who will turn 5 in the 3 months before their state’s kindergarten cutoff, boys are 6 percentage points more likely to redshirt. Among those who turn 5 in the 3 months after the state cutoff, gender is not significantly associated with early kindergarten entry. Associations between demographic characteristics and kindergarten entrance are consistent with the patterns in Table 2.

Table 3

Differences in the Likelihood of Starting Kindergarten in 2006, by Demographic and Developmental Characteristics

	Bivariate models			Multivariate models
	Model	Redshirt	Greenshirt	Redshirt	Greenshirt
Female	1	−0.060*	0.032	−0.057*	0.040^†
		(0.026)	(0.024)	(0.026)	(0.023)
Black	2	−0.136***	0.091*	−0.096***	0.090*
		(0.026)	(0.041)	(0.023)	(0.044)
Hispanic		−0.129***	−0.017	−0.108***	−0.024
		(0.028)	(0.024)	(0.032)	(0.022)
Asian		−0.114**	0.175**	−0.132**	0.220**
		(0.037)	(0.068)	(0.044)	(0.076)
p(F test)		.000	.004	.000	.001
SES	3	0.069***	−0.010	0.058*	−0.013
		(0.018)	(0.014)	(0.023)	(0.016)
Midwest	4	0.069	0.067^†	0.010	0.049
		(0.044)	(0.037)	(0.046)	(0.035)
South		−0.059^†	0.040	−0.063^†	0.028
		(0.030)	(0.027)	(0.035)	(0.025)
p(F test)		.001	.164	.048	.347
Social skills (parents)	5	−0.020	−0.010	−0.007	−0.013
		(0.015)	(0.011)	(0.015)	(0.012)
Social skills (teacher)	6	−0.006	0.002	—	—
		(0.019)	(0.010)
Basic skills (parent)	7	0.009	0.006	−0.008	0.019
		(0.013)	(0.016)	(0.018)	(0.016)
Math	8	0.026^†	0.018^†	0.004	0.015
		(0.014)	(0.010)	(0.022)	(0.015)
Reading	9	0.024^†	0.011	−0.009	0.001
		(0.013)	(0.013)	(0.018)	(0.023)
Low birth weight	10	−0.035	0.025	−0.003	0.045
		(0.026)	(0.032)	(0.027)	(0.033)
Very low birth weight		0.084*	−0.006	0.123**	−0.024
		(0.041)	(0.031)	(0.040)	(0.028)
p(F test)		.027	.698	.007	.205

Note. Results shown in columns 1 and 2 are from 10 separate regression models (linear probability models). The first column shows the coefficient from the model predicting redshirting based on all those born within the 3 months prior to their state’s cutoff date. The second shows the coefficient on the covariate from the model predicting greenshirting based on all those born within the 3 months after their state’s cutoff date. Each model includes a variable indicating the number of months between the child’s birthday and the cutoff date. Excluded groups, by model, are males, White children, none, and western states. The race model includes children who are multiple and other races, but because of small sample sizes, these groups are not shown. For similar reasons, results for the northeastern region are not shown. Columns 3 and 4 show results from multivariate models. The teacher-reported measure of social skills is excluded from these models because it is available only for children who experienced nonrelative care at age 4 and therefore leads to a substantial drop in sample size. However, models that include this variable provide very similar results. All analyses are weighted.

†

p < .1. *p < .05. **p < .01. ***p < .001.

In the lower panel we present results from models that examine whether kindergarten entry decisions are also systematically related to measures of child development collected during the year ECLS-B children turned 4. We might expect that parents of children with lower skills would be more likely to delay their child’s kindergarten entry, and similarly that children with stronger skills would enter early. In general, we do not observe these patterns. We find no relationship between redshirting or early entrance and children’s social skills as measured by either their parents or their teachers. Similarly, there is no relationship between entry behaviors and parents’ assessment of their children’s basic skills. We do observe a weak association between entrance behaviors and children’s direct mathematics and reading assessment: Children with higher preliteracy and math scores are slightly more likely to redshirt than those with lower scores, although this relationship is only marginally significant (p < .10).

Finally, children who had very low birth weights are more likely to delay kindergarten. In additional models, not shown, we found no relationship between kindergarten entry timing and their age 4 height, weight, or body mass index. We therefore posit that the significant relationship between very low birth weight and kindergarten entrance may be capturing a host of developmental outcomes rather than a simple measure of physical development. With the exception of very low birth weight and the “counterintuitive” (and only marginally significant) mathematics and reading results, our measures of child development are not significant predictors of redshirting or early entrance.

Columns 3 and 4 report the coefficients from multivariate regression models that include all the demographic and developmental variables within a single model. The gender, race, and socioeconomic differences in redshirting persist mostly unchanged in the multivariate models controlling for developmental characteristics. The results remain largely unchanged if we also account for family structure, poverty, and other family characteristics (results not shown). Once we account for SES and other demographic covariates, children’s math and reading ability are no longer related to their likelihood of starting kindergarten on time. We still observe that children with very low birth weight are more likely to delay kindergarten; however, no other measures of child development are related to the entry decision.¹⁰

Redshirting and school context

Next we explore whether there is local variation in redshirting rates. Recall that to do this we must turn to an earlier (fall 1998) cohort of kindergarteners. As noted above, we find that in 1998 approximately 5.5% of first-time kindergarteners redshirted. There is a great deal of local variation in redshirting, including schools with no redshirters and schools where more than 25% of sampled children are classified as delayed kindergarten entrants. As shown in Table 4, school-level racial and family-income composition are strongly predictive of redshirting rates, with higher rates in schools serving greater proportions of high-income and White students. For instance, as shown in Figure 5, in schools serving students whose mean household income is $20,000, median redshirting rates were approximately 4% compared to a median redshirting rate of 7% in schools serving students with a mean household income of $100,000. One fifth of schools serving high-income students had redshirting rates of 15% or greater (if most redshirting students were born in the 3 months prior to the state cutoff date, this implies that as many as 60% of those born within 3 months of the cutoff date were redshirters).¹¹

Table 4

School-Level Predictors of Redshirting Practices, 1998 Kindergarten Cohort (Early Childhood Longitudinal Study–Kindergarten Cohort)

	1	2	3
Intercept	−3.273***	−3.285***	−3.290***
	(0.062)	(0.063)	(0.063)
School mean HH income (in 10,000s)	0.071**		0.047^†
	(0.022)		(0.027)
School proportion Black		−0.477	−0.232
		(0.297)	(0.330)
School proportion Hispanic		−1.258***	−0.951*
		(0.360)	(0.401)
School proportion other race		−1.216^†	−1.219^†
		(0.626)	(0.630)
Between-school variance of intercepts	0.966	0.963	0.962
Number of students	12,100	12,100	12,100
Number of schools	685	685	685

Note. Reported coefficients are log odds estimated using random-intercept logistic regression models. Sample restricted to public schools.

†

p < .1. *p < .05. **p < .01. ***p < .001.

Figure 5.

Cumulative distribution of redshirting rates among schools, by school mean household income, 1998.

Implications for Cohort Composition

In this final results section we discuss the implications of redshirting for the within-school distribution of age and cognitive ability at school entry. Using data from the ECLS-K, presented in Table 5, we find that redshirting substantially expands the age distribution of kindergarteners within a school: A 10-percentage-point increase in the proportion of redshirters widens the 90th/10th percentile age range within a school by about 3 weeks. It is somewhat surprising, however, that we find no evidence of an association between within-school redshirting patterns and the distribution of cognitive measures assessed at school entry. Our point estimates suggest that a 10-percentage-point difference in the redshirting rate is associated with a roughly 0.001 difference in the within-school standard deviation of math or reading scores—a trivial, and statistically insignificant, association. Because these estimates are based on cross-sectional associations, however, they should not be interpreted as conclusive evidence that redshirting has no effect of the variation of test scores within school-based kindergarten cohorts.

Table 5

Redshirting and the Standard Deviation of School-Level Developmental Outcomes at School Entry (Early Childhood Longitudinal Study–Kindergarten Cohort)

	Age range (in years)		School standard deviation, reading		School standard deviation, math
School proportion redshirting	0.577***	0.562***	0.009	−0.022	0.012	0.021
	(0.107)	(0.109)	(0.132)	(0.132)	(0.118)	(0.120)
School mean HH income (in 10,000s)		0		0.006		−0.002
		(0.003)		(0.004)		(0.003)
School proportion Black		−0.028		−0.031		−0.046
		(0.032)		(0.039)		(0.036)
School proportion Hispanic		−0.018		−0.049		−0.011
		(0.036)		(0.044)		(0.040)
School proportion other race		−0.048		0.159*		0.083
		(0.052)		(0.063)		(0.057)
Constant	0.767***	0.782***	0.865***	0.836***	0.867***	0.877***
	(0.008)	(0.025)	(0.010)	(0.030)	(0.009)	(0.028)
N (schools)	640	640	640	640	640	640

Note. Age range is defined as the difference between the 90th and 10th percentiles of birthdays. All models include state fixed effects and are restricted to public schools with 10 students or more.

p < .05. ***p < .001.

Discussion

Our investigation of delayed kindergarten entrance adds to the literature on school entry in several ways. First, we leverage a large, nationally representative data set to provide a detailed exploration of families’ decisions to delay kindergarten entry. Unlike previous studies that observed children only once they entered kindergarten or later in life, our study makes use of data on a full age cohort of 5-year-olds, some of whom did and did not enter kindergarten in 2006. In addition, we have access to parent, caregiver, and direct assessments of children’s development prior to age 5, so we can carefully examine whether parents’ decisions about kindergarten entry are related to observable measures of child development.

Our findings show that nationwide approximately 4% to 5.5% of children delay kindergarten, a rate lower than those previously reported. We demonstrate that the likelihood of redshirting is much higher among children whose birthday falls in the months before the cutoff and that redshirting varies substantially across gender, race, and SES. Poor families rarely delay kindergarten despite the fact that they are far more likely to indicate concerns about their children’s school readiness at age 4. The findings are consistent with our initial hypothesis that low-income families may view redshirting as prohibitively expensive. Giannarelli and Barsimantov (2000) report that, on average, low-income families spend 16% of their earnings on child care. Within this context, the patterns we observe are not surprising.

We also examine whether rates of redshirting, as well as the gaps in these rates across groups, are explained in part by observable measures of child development including parent and teacher reports and direct assessments. We find little evidence that this is the case. In fact, children who redshirt tend to perform slightly better during prekindergarten than many of their similarly aged complier peers, although these differences are eliminated if we account for SES. The absence of a relationship between nearly all of the developmental measures and the likelihood of delayed kindergarten entry suggests parents’ decisions to redshirt may be driven more by concerns about their children’s relative age or relative physical development in their classroom than by their actual cognitive or behavioral development. The evidence on the impact of relative age on student learning trajectories is mixed, and recent research (Black, Devereux, & Salvanes, 2008; Deming & Dynarski, 2008; Dobkin & Ferreira, 2010) suggests benefits from delayed school admission may be counteracted by losses associated with one less year in the labor force. Still, parents’ desire to give their child an edge in what they perceive as a competitive environment may make it appealing for those who can afford it to give their child the “gift of time” (Gesell, 1940).

Given the differences in redshirting practices across groups, it seems plausible that redshirting would lead, mechanically, to exacerbated achievement gaps. In related analyses using the ECLS-B data, we find that although redshirting substantially affects an individual child’s initial standing within a kindergarten cohort, it does not meaningfully widen achievement gaps (Bassok & Reardon, 2012). This is because, at the national level, the incidence of redshirting is low enough that its impact is not large. However, this does not imply that redshirting bestows benefits for selected individuals without causing any negative repercussions for others. It is plausible that in certain communities where redshirting levels are high there are meaningful implications for learning experiences as well as achievement.

The ECLS-K data allowed us to measure, first, to what extent redshirting rates vary locally and, second, whether within-school redshirting rates are related to differences in the composition of kindergarten cohorts. We find that indeed redshirting rates vary substantially between schools and that the racial and socioeconomic composition of schools is strongly related to redshirting. In future work it would be worthwhile to explore how school-level redshirting norms are formed and to what extent kindergarten teachers (or preschool teachers in the communities surrounding these schools) transmit these norms to parents through their advice about school-entry timing.

Finally, we find a strong relationship between schools’ redshirting rates and the range of ages among the kindergarteners within a school, which in turn could translate to meaningful differences in learning experiences. Unfortunately, we do not have data that allow us to explore that hypothesis. We do show, however, that at the school level, redshirting is not related to meaningful differences in the distribution of cognitive assessment scores at school entry. These admittedly crude results show that even at the school level, redshirting is not associated with meaningful changes in cohort composition, at least with respect to the assessment scores we considered. Taken together then, the findings suggest that the policy implications of current redshirting practices are more limited than media reports suggest.

It is worth noting that our study, and particularly the within-school analysis, does suffer from several data limitations. In particular, the within-school analysis does not measure the effect of redshirting on school-level composition. Unlike the ECLS-B, which allows us to observe redshirters in both the year they were eligible for kindergarten and the year they enrolled, in the ECLS-K we observe a single snapshot of a kindergarten cohort. This means that although we can measure the association between schools’ redshirting rates and various characteristics of their incoming kindergarten cohorts, we have no way to compare that to a counterfactual scenario in which redshirters started on time. Furthermore, the children in the ECLS-K started kindergarten nearly 15 years ago, in a period that preceded large state preschool expansions as well as No Child Left Behind. Although our analysis provides a detailed account of which types of children redshirt in more recent years, it is difficult to know whether similar patterns existed in 1998. Soon-to-be-released ECLS data from 2010 will allow researchers to assess changes over time in within-school patterns of redshirting.

Footnotes

Notes

Authors

DAPHNA BASSOK is Assistant Professor of Education at the University of Virginia, 405 Emmet Street South, PO Box 400277, Charlottesville, VA 22904; db9ec@virginia.edu. Her research focuses on early childhood education policy.

SEAN F. REARDON is Professor of Education at Stanford University. His research focuses on education policy and inequality.

References

Barnett

W. S.

Epstein

D. J.

Friedman

A. H.

Sansanelli

Hustedt

J. T.

(2009). The state of preschool, 2009. New Brunswick, NJ: National Institute for Early Education Research.

Bassok

Reardon

(2012). Has redshirting changed kindergarten? Evidence from synthetic cohorts. Working paper.

Bazelon

(2008, August 1). The downside of redshirting: The trouble with older kindergarten. Slate. Retrieved from http://www.slate.com/id/2196423/

Black

S. E.

Devereux

P. J.

Salvanes

K. G.

(2008). Too young to leave the nest: The effects of school starting age (NBER working paper). Cambridge, MA: National Bureau of Economic Research.

Byrd

R. S.

Weitzman

Auinger

(1997). Increased behavior problems associated with delayed school entry and delayed school progress. Pediatrics, 100(4), 654–661. doi:10.1542/peds.100.4.654

Cosden

Zimmer

Tuss

(1993). The impact of age, sex, and ethnicity on kindergarten entry and retention decisions. Educational Evaluation and Policy Analysis, 15(2), 209–222.

Cuban

(1992). Why some reforms last: The case of the kindergarten. American Journal of Education, 100(2), 166–194.

Datar

(2006a). Does delaying kindergarten entrance give children a head start? Economics of Education Review, 25(1), 43–62.

Datar

(2006b). The impact of kindergarten entrance age policies on the childcare needs of families. Journal of Policy Analysis and Management, 25(1), 129–153.

10.

Deming

Dynarski

(2008). The lengthening of childhood. Journal of Economic Perspectives, 22(3), 71–92.

11.

Dobkin

Ferreira

(2010). Do school entry laws affect educational attainment and labor market outcomes? Economics of Education Review, 29(1), 40–54.

12.

Frey

(2005). Retention, social promotion, and academic redshirting. Remedial and Special Education, 26(6), 332–346.

13.

Gesell

(1940). The first five years of life: A guide to the study of the preschool child, from the Yale Clinic of Child Development. New York, NY: Harper & Bros. Retrieved from http://www.getcited.org/pub/101170878

14.

Giannarelli

Barsimantov

(2000). Child care expenses of America’s families. Washington, DC: Urban Institute.

15.

Graue

M. E.

DiPerna

(2000). Redshirting and early retention: Who gets the “gift of time” and what are its outcomes? American Educational Research Journal, 37, 509–534.

16.

Graue

M. E.

Kroeger

Brown

(2002). Living the “gift of time.” Contemporary Issues in Early Childhood, 3(3), 338–353.

17.

Jacob

B. A.

(2005). Accountability, incentives and behavior: The impact of high-stakes testing in the Chicago Public Schools. Journal of Public Economics, 89(5–6), 761–796.

18.

Lincove

J. A.

Painter

(2006). Does the age that children start kindergarten matter? Evidence of long-term educational and social outcomes. Educational Evaluation and Policy Analysis, 28(2), 153–179.

19.

Matlack

(2011, May 15). Redshirting kindergarten: Why are so many parents holding their boys back? Is it really good for them? And what impact is it having on everybody else? Huffington Post. Retrieved from http://www.huffingtonpost.com/tom-matlack/redshirting-kindergarten_1_b_859824.html

20.

Noel

A. M.

Newman

(2003). Why delay kindergarten entry? A qualitative study of mothers’ decisions. Early Education & Development, 14(4), 479–498.

21.

O’Donnell

Mulligan

(2008). Parents’ reports of the school readiness of young children from the National Household Education Surveys Program of 2007: First look. Washington, DC: National Center for Education Statistics.

22.

Paul

(2010, August 20). The littlest redshirts sit out kindergarten. New York Times. Retrieved from http://www.nytimes.com/2010/08/22/fashion/22Cultural.html?pagewanted=all

23.

Shonkoff

J. P.

Phillips

(2000). From neurons to neighborhoods: The science of early childhood development. Washington, DC: National Academy Press.

24.

Stipek

(2002). At what age should children enter kindergarten? A question for policy makers and parents. Social Policy Report, 16(2), 3–16.

25.

U.S. Department of Commerce. (2006). Current Population Survey, October Supplement. Washington, DC: Author.

26.

Weil

(2007, June 3). When should a kid start kindergarten? New York Times. Retrieved from http://www.nytimes.com/2007/06/03/magazine/03kindergarten-t.html?pagewanted=all

27.

Zill

Loomis

L. S.

West

(1997). The Elementary school performance and adjustment of children who enter kindergarten late or repeat kindergarten: Findings From national surveys. Washington, DC: National Center for Education Statistics. Retrieved from http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=98097