USFDC Home   RSS 
Material Information
Subjects
Record Information

Full Text 
PAGE 1 EDUCATION POLICY ANALYSIS ARCHIVES A peerreviewed scholarly journal Editor: Gene V Glass College of Education Arizona State University Copyright is retained by the first or sole author who grants right of first publication to the Education Policy Analysis Archives EPAA is a project of the Education Policy Studies Laboratory. Articles are indexed in the Dir ectory of Open Access Journals (www.doaj.org). Volume 12 Number 49 September 18, 2004 ISSN 10682341 Age and Achievement* James B. Grissom California Department of Education Citation: Grissom, J. B. (2004, September 18). Age and achievement. Education Policy Analysis Archives, 12(49). Retrieved [date] from http://epaa.as u.edu/epaa/v12n49/. Abstract There is continuing controversy abou t the optimal or appropriate age at which children should start school. The purpose of this study is to examine the relationship between age and achievem ent. It is an attempt to evaluate the hypothesis that older students fare better academically than their younger classmates. Findings indicate that on average for students in elementary school there is positive linear rela tionship between age and achievement for age normal peers Even though there is positive linear relationship, the difference in average test scores between the oldest and youngest students is not great and by the time students reach 10th grade the positive linear relationship has disappeared. For overage students there is on average a negative linear relationship between age and achievement at all grade levels. That is, the negative relationship between age and achievement remains constant over time. These results argue against modifying entrance age policies, delaying school entry, implemen ting transitional kindergarten or first grade programs or retaining students to improve educational achievement. Policies and practices that make students older than their classmates inversely affect their educational achievement. *The opinions expressed are of the author alone and do not reflect opinion or policy of the California Department of Education. There is continuing controversy about the optimal or appropriate age at which children should start school. The efficacy of delaying sc hool entry beyond the ag e that a student can legally enroll in public school has been debated in the research literature. Crosser (1991), PAGE 2 Age and Achievement 2 Kinard & Reinherz (1986), and La Paro & Pianta (2000) present evidence that older children fare better academically than their younger, age appropriate peers. Uphoff & Gilmore (1985) use research evidence about the relationship between age and achievement as well as other evidence to argue that the older and/or more mature students in a class fare better than younger classmates. In contrast DeMeis & S tearns (1992) and Dietz & Wilson (1985) found no significant relationship between age and ac hievement. Langer, Kalk, & Searls (1984) found significantly higher achievement of the olde st as compared to the youngest students at age nine but this difference disappeared by age seventeen. Shepard (1997) argues that even if more emotionally mature children do better in school, there are no valid instruments or means to identify these children. The most popular readiness tests (e.g., Lights Retention Scale, the Gesell School Readiness Test, th e Gesell Preschool Test, the Brigance K & 1 Screen, the Daberon Screening for School Readin ess, the Developmental Indicators for the Assessment of LearningRevised, and th e Missouri Kindergarten Inventory of Developmental Skills) all lack demonstrated validity and reliability to make readiness decisions about individual children. Meisels (199 2) argues that when parents practice delayed school entry, younger, legally enrolled students may be disadvantaged. When first graders who are barely 6 years old are compared to 7 year olds, the 6 year olds, functioning at a developmentally appropriate level, seem immatu re. Immaturity is one of the major reasons given for grade retention (Abidin, Golladay, & Howerton, 1971; Niklason, 1984). The controversy continues despite the fact that no matter where the legal entry age is set there will be older and younger children (i.e., re lative to each other) in a class. The incidence of delayed school entry and transi tional kindergarten and first grade programs has been fueled by the belief among parents and educators that the curriculum is too difficult (Shepard & Smith, 1988). The academic expectations at each grade level have been raised and so the curriculum is being pushed down. That is, what was once expected of students in first grade is now expected of st udents in kindergarten. As a means to protect children from the more demanding curriculum some educators recommend (Uphoff & Gilmore, 1985) and some parents practice delayed school entry (Graue & DiPerna, 2000). It is thought that older more mature children are better prepared developmentally (i.e., both cognitively and emotionally) to handle the more rigorous curriculum. In the current climate of accountability teach ers are being held responsible for student outcomes on standardized tests. Teachers are also responsible for making sure students have mastered the material in the current grade so that students are prepared to master material in the next grade. The result is that teachers want more homogenous and cognitively advanced classrooms so that the outcomes for which they are held accountable can be more efficiently produced (Smith & Shepard, 1987). In their own selfinterest, teachers want classrooms of students that will be successful on standardized tests. Teachers therefore favor policies and practices (e.g., delayed school entry, changing the legal age of entry, and transitional kindergarten programs) they believe produce such classrooms. The negative education consequences of homogeneous grouping and tracking have b een described in the research literature. Homogeneous ability grouping enhances the achievement ability of the fast group and retards the achievement of the slower group (Smith & Shepard, 1987). These consequences run counter to the democratic goals of education in the United St ates. In addition, as long as classrooms are made up of 20 to 30+ students, efforts to re duce heterogeneity will largely be futile. There will always be variability in terms of educational achievement. PAGE 3 Education Policy Analysis Archives Vol. 12 No. 49 3 There is a strong belief among parents and educators that grade retention allows children time to mature cognitively to handle the more rigorous curriculum (Byrnes, 1989; Combs & Tanner, 1993; Smith, 1989). It is also seen by teachers as another way to reduce heterogeneity. In spite of these beliefs, resear ch on grade retention indicates that retained students do less well academically when co mpared to recommended but not retained students (Holmes, 1989; Holmes & Matthews, 1984). Even so, grade retention remains an accepted academic intervention to raise studen t achievement. One certain effect of grade retention is that it makes students older than their grade level peers. Angrist & Krueger (1992) tested the hypothesi s that there is inverse relationship between educational attainment and age at school entr y. That is, students who enter school at an older age drop out after having completed less schooling (i.e., because they are legally able to do so) than students who enter school at a younger age. The argument by Angrist & Krueger (1992) is that leaving early reduces the number of years of schooling and thus educational attainment. Grissom & Shepard (1989) present evidence that retained and/or overage students are more likely to drop out of school than students not retained and/or not overage after controlling for achievement diffe rences. Policies and practices that make students older than their classmates increase the likelihood that these students will leave school early. Most students will be older than their age normal peers for three reasons: they started school late, spent two years in a transitional kindergar ten or first grade prog ram, or were flunked and forced to repeat a grade. The belief remains strong that these three academic interventions benefit students academically and in other ways, despite evidence to the contrary. The belief is so strong that there are laws that mandate grade retention as the preferred remediation for lowachieving studen ts. For example, California Education Code, Section 48070.5 (d) states that, If a pupil is performing below the minimum standard for promotion, the pupil shall be retained in his or her current gr ade level unless the pupil's regular classroom teacher determines in writing that retenti on is not the appropriate intervention for the pupil's academic deficiencies. The major purpose of this study is to examine the relationship between age and achievement. It evaluates the hypothesis that older students fare better academically (e.g., score higher on standardized tests) than their younger classmates and it also evaluates the hypothesis that children are protected (i.e., fr om an unrealistic and harsh curriculum) and benefit from delayed school entry, tran sitional programs, and/or grade retention. A secondary purpose is to present some evidence on the extent of academic redshirting in California public schools. Researchers sometime s refer to the practice of delaying school entry as academic redshirting. Although available data cannot identify why students are overage, an examination of age dist ributions may allow some inferences. Researchers (Brent, May, & Kundert, 1996; Br acy, 1989) have reported on an increasing trend to delay school entry for ageeligible children. Though Bellisimo, Sacks, & Mergendoller, (1995) reported a drop in this trend in California between 1989 and 1991, PAGE 4 Age and Achievement 4 there is reason to believe that that academic r edshirting is on the rise due to the increasing curriculum demands of kindergarten and first grade (Graue & DiPerna, 2000). Researchers also indicate that males are more likely to ex perience delayed school entry than females (Bellisimo, Sacks, & Mergendoller, 1995; Bren t, May, & Kundert, 1996; Graue & DiPerna, 2000; May & Kundert, 1995) and that parents identified as having higher socioeconomic status (SES) are more likely to delay school entr y than those parents identified as lower SES (Bellisimo, Sacks, & Mergendoller, 1995). Most studies of entrance age practices are based on data collected at the district level, which limits the ability to generalize to larger popu lations. In contrast, Langer, Kalk, & Searls, (1984) had nationally representative data (i.e., NAEP) but had to infer school entry practices based on student age and school entry policies. Graue & DiPerna (2000) addressed weaknesses of earlier studies by using sample se lection strategies that allowed them to collect data as to when students started school an d make inferences at the state level (i.e., Wisconsin). This study is based on data collected on stud ents enrolled in California public schools in grades 2 through 11 and suffers the limitations of Langer, Kalk, & Searls, (1984). That is, data were not collected until second grade and so entrance age practices have to be inferred from age distributions and st atewide entrance age policies. Method Each spring California administers a series of standardized achievement tests known as the Standardized Testing and Reporting (STAR) prog ram. Tests are administered to all public school students enrolled in grades 2 through 11. As part of the testing program, demographic information, including birth date, is collected. STAR tests were administered first in the spring of 1998. From 1998 to 2002 a normreferenced standardized test, the Stanford Achievement Test version 9 (SAT/9) form T, was administered as part of the STAR program. This study uses data from tests administered in the spring of 1998 through 2002. Students in California need to be five years old by December 2 to enroll in kindergarten or six years old by December 2 to enroll in first grade. For students tested in 2002, student age on December 1, 2001 (i.e., the December before spring testing) was calculated in months. As stated, the youngest students tested in spri ng 2002 were second gr aders. The youngest second graders were students who turned 7 close to the cutoff date of December 2. The age of the youngest second graders was 85 months1. Second grade students who were 85 months 1 Students who were 84 12 7 months old were students who turned seven on December 1. Although these were the youngest students in the cohort, they were not included in the analyses. The study concerns test scores by month. The youngest students born in December would only include students born on one day (i.e., December 1). The number of second grade students born on December 1, 1994 and tested in spring 2002 was 856. The total number of students tested in spring 2002 was 485,796. Including this small number as a unique category is not helpful in making overall inferences. Adding students born on one day in December to those born in November is not going to change the general relationship PAGE 5 Education Policy Analysis Archives Vol. 12 No. 49 5 old were students with November 1994 birt h dates. Second grade students who were 86 months old were students with October 1994 birth dates and so on. Second grade students who were 96 months old were students with December 1993 birth dates. Students with December 1993 birth dates were the oldest age normal peers2. Second grade students who were 97 months or older had been retained. Student demographic information does not contain info rmation as to why students were held back. That is, there is no information as to whether students started school late (i.e., academic redshirting), spent two years in a transitiona l kindergarten or first grade program, or were retained in grade (i.e., flunked) in kindergarten or first grade. Students who were 97 months old have November 1993 birth dates. These were the retained (i.e., held back for one of the three stated reasons) students whos e birth dates are closest to the cutoff date. Students who were 98 months old were the retained students with October 1993 birth dates and so on. For each age (in months) the average SAT/9 total reading and mathematics scores were calculated in normal curve equivalent (NCE) unit s. The average test score by age provides an indicator of the relationship between age and achievement. Given the errors in selfreport data and the desire to avoid discussion about students who are very young or very old relative to thei r agenormal peer group, the full age range was truncated. For second grade students the age range was truncated to students who were 85 to109 months old. The oldest students are 24 months or two years older than the youngest students. Results Age and Achievement Figure 1 shows the SAT/9 average total reading NCE score for students in grade 2 by age in months who were tested in spring 2002. between age and achievement. In fact, the av erage normal curve equivalent (NCE) score for students born on December 1, 1994 is 48.5. Th e average NCE score for students born in November 1994 is 48.4. Since these values are e ssentially the same, it was easier to exclude them from the analyses than determine how to include them. 2 Normal age peers are students who start school as soon as they are legally able. PAGE 6 Age and Achievement 6 Fi g ure 1. SAT/9 mean total readin g score b y a g e in months: Grade 2 STAR 2002, n = 455,638 20 25 30 35 40 45 50 55 60 65 70 75Nov94 Oct94 Sep94 Aug94 Jul94 Jun94 May94 Apr94 Mar94 Feb94 Jan94 Dec93 Nov93 Oct93 Sep93 Aug93 Jul93 Jun93 May93 Apr93 Mar93 Feb93 Jan93 Dec92 Nov92858687888990919293949596979899100101102103104105106107108109 Birth Month & Year Age in MonthsSAT/9 Mean Reading NCE Score The first age in figure 1 (i.e., 85 months ) shows the mean total reading NCE score for students with November 1994 birthdays. These are the youngest students in this particular cohort. The second age (i.e., 86 months) shows the mean total reading NCE score for students with October 1994 birthdays. The age normal peer group for this 2nd grade cohort ranges from 85 months to 96 months. There is a positive relationship between age an d achievement for the age normal peers. As age normal peers get older, their tes t scores on average get higher. For students who have been retained (i.e., stud ents who are 97 months and older) there is a negative relationship between age and achievemen t. As students get older, their test scores on average decline. PAGE 7 Education Policy Analysis Archives Vol. 12 No. 49 7 Next, data were analyzed to determine whether the relationship between age and achievement is content dependent. Figure 2 shows the SAT/9 average total mathematics NCE score for students in grade 2 by age in months who were tested in spring 2002. Fi g ure 2. SAT/9 mean total mathematics score b y a g e in months: Grade 2 STAR 2002, n = 469,805 20 25 30 35 40 45 50 55 60 65 70 75Nov94 Oct94 Sep94 Aug94 Jul94 Jun94 May94 Apr94 Mar94 Feb94 Jan94 Dec93 Nov93 Oct93 Sep93 Aug93 Jul93 Jun93 May93 Apr93 Mar93 Feb93 Jan93 Dec92 Nov92858687888990919293949596979899100101102103104105106107108109 Birth Month & Year Age in MonthsSAT/9 Mean Mathematics NCE Scores The data pattern in figure 2 is similar to that of figure 1 with a positive relationship between age and achievement for age normal peers and a negative relationship for students who have been retained. The relationship between age and achievement is not content dependent. To test the linear relationship between age and achievement, SAT/9 mean NCE total reading scores were regressed on age in months. First, the relationship was tested for the age normal peers. Table 1 shows these results. Table 1 PAGE 8 Age and Achievement 8 SAT/9 mean total reading NCE score regressed on age in months for grade 2 age normal peers ANOVA df SS MS F Significance F RSquare Regression 1 38.3973 38.3973 293.2372 0.00000 0.9670 Residual 10 1.3094 0.1309 Total 11 39.7067 There is a statistically significant relationshi p between age and mean achievement for the age normal peer group and the relationship is strong (i.e., the 2 R = .97). Figure 3 graphically displays the relationship between mean test scor es and age for the age normal peer group. Figure 3 also displays the regression equation (i.e., )) 5182 0 ( 7395 4^ agex y For each month age increases, the average total reading NCE score increases point. Fi g ure 3. SAT/9 mean total readin g NCE score re g ressed on a g e for the age normal peer group: Grade 2 STAR 2002 y = 0.5182x + 4.7395 R2 = 0.967 47 48 49 50 51 52 53 54 55 8485868788899091929394959697 Age in MonthsSAT/9 Mean Reading NCE Score PAGE 9 Education Policy Analysis Archives Vol. 12 No. 49 9 Figure 3 shows a positive linear relationship between mean test score and age for the age normal peer group. Next, SAT/9 total reading scores were regresse d on age in months to test the relationship between age and achievement for retained st udents. The ages of 97 months and higher represent the students who had been held back. As stated, the student demographic information does not contain information as to why students were held back. Table 2 shows these results. Table 2 SAT/9 mean total reading NCE score regressed on age in months for grade 2 retained students ANOVA df SS MS F Significance F RSquare Regression 1 190.456 190.456 87.71862 0.0000014 0.888572 Residual 11 23.8834 2.17122 Total 12 214.339 There is a statistically significant negative relationship between age and achievement for retained students and the relationship is strong (i.e., 2 R = .89). Figure 4 graphically displays the relationship between mean test scores and age for retained students and the regression equation (i.e., )) 023 1 ( 84 146^ agex y. For each month age increases, the average total reading NCE score decreases 1 point. PAGE 10 Age and Achievement 10 Fi g ure 4. SAT/9 mean total readin g NCE score re g ressed on a g e for retained students: Grade 2 STAR 2002 y = 1.023x + 146.84 R2 = 0.8886 32 34 36 38 40 42 44 46 48 50 52 54 56 96979899100101102103104105106107108109110 Age in MonthsSAT/9 Mean Reading NCE Score Figure 4 shows that test scores begin to declin e for retained students and continue to do so through the age range. The variance of mean test scores for retained students is approximately 16 points. That is, even thought the 2 R is less for grade 2 retained students than for age normal peers, the difference in test scores for retained students is almost three times the difference in test scores for age normal peers. Educators and researchers that recommend dela yed school entry for students typically mean the students who are one to three months older than the age normal peer group. The summer birth date research concerns states where the cut off for school entry is around September 1. Students with summer birth dates are those students with birth dates three months prior to a September 1 cut off. In Califo rnia, students need to be five years old by December 2 to enroll in kindergarten or six years old by December 2 to enroll in first grade. Therefore, the relationship between age and achi evement was tested when students who are PAGE 11 Education Policy Analysis Archives Vol. 12 No. 49 11 one to three months older than the age nor mal peer group were added to the age normal peer group. Table 3 shows these results. Table 3 SAT/9 mean total reading NCE score regressed on age in months for grade 2 normal ag e peers and retained students ANOVA df SS MS F Significance F R Square Regression 1 0.1877 0.1877 0.0326 0.8594 0.0025 Residual 13 74.7667 5.7513 Total 14 74.9544 There is no longer a statistically significant linear relationship between age and achievement. The 2 R = .003. When the sample includes retained students, the positive linear relationship between age and achievement disappears. Figure 5 graphically displays these same data. Figure 5. SAT/9 mean total reading NCE score regressed on age for the age normal peer group plus three months overage: Grade 2 STAR 2002 y = 0.0259x + 48.546 R2 = 0.0025 45 46 47 48 49 50 51 52 53 54 55 84858687888990919293949596979899100 Age in MonthsSAT/9 Mean Reading NCE Score PAGE 12 Age and Achievement 12 Being older is better to a point. Beyond that point the effect is negative. Age normal students who are older do better on average. However students who are older because they have been retained do worse on average. The reco mmendation to delay school entry or retain students to improve academic achievement is not supported by these data. Examining mean test scores provides a simplified way to examine the relationship between age and achievement. However, using the mean al so disguises the actual relationship. Table 4 shows the results of regressing total reading scor es for individual students on age in months. Table 4 SAT/9 total reading NCE sc ore regressed on age in months for grade 2 age normal peers ANOVA df SS MS F Significance F R Square Regression 1 993813 993813 2695.52<.0001 0.0069 Residual 390251 143881916 368.69070 Total 390252 144875728 The positive relationship between age and test scor es is still statistically significant. However, the 2 R = .0069. That is, even though there is a statistically significant relationship between age and achievement, age accounts for little of the variance in test scores. The regression equation is: 473 743 8agex y Figure 6 graphically displays these data. PAGE 13 Education Policy Analysis Archives Vol. 12 No. 49 13 These data indicate that making strong infere nces about student academic performance if age is known is not sound. Many students with November 1994 birth dates (i.e., the youngest students) performed well and many students with December 1993 birth dates (i.e., the oldest students) performed poorly. Table 5 and figure 7 show the relationship between age and achievement for retained students when total reading scores for individual students are regressed on age. Table 5 SAT/9 total reading NCE sc ore regressed on age in months for grade 2 retained students ANOVA df SS MS F Significance F R Square Regression 1 1065911 1065911 2980.03 <.0001 0.0436 Residual 65383 23386492 357.68460 Total 65384 24452403 The regression equation is: 254 1 413 170 agex y. The negative relationship between age and test scores is still statistically signif icant. Again, age accounted for little of the PAGE 14 Age and Achievement 14 variance in test scores (i.e., 2 R = .0436) but it accounted for more of the variance than it did for the age normal peer group. This may be due to the fact that there were fewer students in the retained group than in the age normal peer group. Fewer students mean less variance and so age has less variance for which to account. Or, the negative relati onship between age and achievement is stronger for retained students than the positive relationship between age and achievement for the age normal peers. To determine if the relationship between age an d achievement is maintained over time, the relationship between age and achievement was tested for older students (i.e., grade 6 students). Figure 8 shows the average total re ading NCE scores for students in grade 6 by age in months who were tested in spring 2002. Fi g ure 8. SAT/9 mean total readin g score b y a g e in months: Grade 6 STAR 2002, n= 465,633 20 25 30 35 40 45 50 55 60 65 70 75Nov90 Oct90 Sep90 Aug90 Jul90 Jun90 May90 Apr90 Mar90 Feb90 Jan90 Dec89 Nov89 Oct89 Sep89 Aug89 Jul89 Jun89 May89 Apr89 Mar89 Feb89 Jan89 Dec88 Nov88133134135136137138139140141142143144145146147148149150151152153154155156157 Birth Month & Year Age in MonthsSAT/9 Mean Reading NCE Scores The first age in figure 8 (i.e., 133 months) shows the mean NCE total reading score for students with November 1990 birth dates. These are the youngest students in this particular cohort. The second age (i.e., 134 months) shows the mean NCE total reading score for PAGE 15 Education Policy Analysis Archives Vol. 12 No. 49 15 students with October 1990 birth dates. The age normal peer group for this cohort ranges from 133 to 144 months. Retained students are 145 months and older. As with grade 2 students there is a positiv e relationship between age and achievement for age normal peers (i.e., students 133 to 144 m onths old). As students get older their test scores get higher. For retained students (i.e., students 145 to 157 months old) there is a negative relationship between age and achievement. As students get older their test scores get lower. The relationship between age and achievement is consistent across two grades. Again data were analyzed to determine whether the relationship between age and achievement is content dependent. Figure 9 shows the SAT/9 average total mathematics NCE scores for students in grade 6 by age in months who were tested in spring 2002. Fi g ure 9. SAT/9 mean total mathematics score b y a g e in months: Grade 6 STAR 2002, n = 469,486 20 25 30 35 40 45 50 55 60 65 70 75Nov90 Oct90 Sep90 Aug90 Jul90 Jun90 May90 Apr90 Mar90 Feb90 Jan90 Dec89 Nov89 Oct89 Sep89 Aug89 Jul89 Jun89 May89 Apr89 Mar89 Feb89 Jan89 Dec88 Nov88133134135136137138139140141142143144145146147148149150151152153154155156157 Birth Month & Year Age in MonthsSAT/9 Mean Mathematics NCE Score PAGE 16 Age and Achievement 16 The data in Figure 9 are similar to that of fi gure 8. That is, there is a positive relationship between age and achievement for age normal peer s and a negative relationship for students who have been retained. The relationship between age and achievement is not content dependent. To test the significance of the relationship b etween age and achievement, mean total reading NCE score were regressed on age in months. Table 6 shows these results for age normal peers. Table 6 SAT/9 mean total reading NCE score regressed on age in months for grade 6 age normal peers ANOVA df SS MS F Significance F R Square Regression 1 26.0252 26.0252 101.8848 0.00000 0.9106 Residual 10 2.5544 0.2554 Total 11 28.5796 Results indicate that there is a statistically significant positive relationship between age and achievement. The 2 R for grade six students (i.e., .91) is lower than the 2 R for grade 2 students (i.e., .97). The positive relations hip between age and achievement for age normal peers decreased in strength as students aged. Figure 10 graphically displays the mean read ing NCE scores for normal peers regressed on age in months. PAGE 17 Education Policy Analysis Archives Vol. 12 No. 49 17 Figure 10. SAT/9 mean total reading NCE score regressed on age for the age normal peer group: Grade 6 STAR 2002 y = 0.4266x 9.6028 R2 = 0.9106 46 47 48 49 50 51 52 53 132133134135136137138139140141142143144145 Age in MonthsSAT/9 Mean Reading NCE Score Figure 10 displays the positive linear relationship between mean test score and age for the age normal peer group. Next, mean total reading NCE scores were regre ssed on age in months for retained students. Table 7 shows these results. Table 7 SAT/9 mean total reading NCE score regressed on age in months for grade 6 retained students ANOVA dfSS MS F Significance F R Square Regression 1 271.9146 271.9146 92.6208 0.00000 0.8938 Residual 11 32.2936 2.9358 Total 12 304.2082 As with grade 2, there is a statistically si gnificant negative relationship between age and achievement. The 2 R for grade six students (i.e., .89) is the same as the 2 R for grade 2 students (i.e., .89). Through grade six the strength of the relationship between age and achievement for retained students has remained constant. PAGE 18 Age and Achievement 18 Figure 11 shows these same results. Fi g ure 11. SAT/9 mean total readin g NCE score re g ressed on a g e for retained students: Grade 6 STAR 2002 y = 1.2223x + 223.95 R2 = 0.8938 30 32 34 36 38 40 42 44 46 48 50 144145146147148149150151152153154155156157158 Age in MonthsSAT/9 Mean Reading NCE Score Figure 11 shows that test scores begin to declin e for retained students and continue to do so through the age range. Next, the relationship between age and achievement was tested for grade 10 students. Figure 12 shows the average total reading scores for students in grade 10 by age in months who were tested in spring 2002. PAGE 19 Education Policy Analysis Archives Vol. 12 No. 49 19 Fi g ure 12. SAT/9 mean total readin g score b y a g e in months: Grade 10 STAR 2002, n=386,910 20 25 30 35 40 45 50 55 60 65 70 75Nov86 Oct86 Sep86 Aug86 Jul86 Jun86 May86 Apr86 Mar86 Feb86 Jan86 Dec85 Nov85 Oct85 Sep85 Aug85 Jul85 Jun85 May85 Apr85 Mar85 Feb85 Jan85 Dec84 Nov84181182183184185186187188189190191192193194195196197198199200201202203204205 Age in MonthsSAT/9 Mean Reading NCE Score The first age in figure 12 (i.e., 181 months) shows the mean NCE total reading score for students with November 1986 birth dates. These are the youngest students in this particular cohort. The age normal peer group for this c ohort ranges from 181 to 192 months. Retained students are 193 months and older. Again, mean total reading NCE score were regressed on age in months to test the significance of the relationship between age an d achievement. Table 8 shows these results for age normal peers. PAGE 20 Age and Achievement 20 Table 8 SAT/9 mean total reading NCE score regressed on age in months for grade 10 age normal peers ANOVA df SS MS F Significance F R Square Regression 1 5.4747 5.4747 13.5335 0.00425 0.5751 Residual 10 4.0453 0.4045 Total 11 9.5200 Results indicate that there is a statistically significant positive linear relationship between age and achievement. However, the2 R for grade ten students (i.e., .58) is relatively low compared to students in grades two and six. Figure 13 shows these same results. Figure 13. SAT/9 mean total reading NCE score regressed on age for the age normal peer group: Grade 10 STAR 2002 y = 0.1957x + 5.7803 R2 = 0.5751 40 41 42 43 44 45 180181182183184185186187188189190191192193 Age in MonthsSAT/9 Mean Reading NCE Score PAGE 21 Education Policy Analysis Archives Vol. 12 No. 49 21 When students are 187 months, achievement le vels off until 190 months. At 191 months, achievement begins to decline. The variance in mean test scores is only a couple of points. Despite statistical significance, it is safe to say that there is no longer a positive linear relationship between age and achievement for ag e normal peers. I make this statement for a couple of reason reasons. The first is that the oldest age normal peers do not have the highest average test scores. Second, the varian ce in test scores for the age normal peers is very small. That is, there is statistical significance but no practical difference in test scores. Whatever academic advantage being older had for younger students is gone by grade ten. Figure 14 shows the results of fitting a 2nd order polynomial through the data. Fi g ure 14. SAT/9 mean total readin g NCE score re g ressed on a g e for the age normal peer group: Grade 10 STAR 2002 y = 0.0527x2 + 19.857x 1827 R2 = 0.9646 40 41 42 43 44 45 180181182183184185186187188189190191192193 Age in MonthsSAT/9 Mean Reading NCE Score These results emphasize that there is no long er a positive linear relationship between age and achievement for age normal peers. Mean total reading NCE scores were regressed on age in months for grade 10 retained students. Table 9 shows these results. PAGE 22 Age and Achievement 22 Table 9 SAT/9 mean total reading NCE score regressed on age in months for grade 10 retained students ANOVA df SS MS F Significance F R Square Regression 1 219.4507 219.4507 161.3951 0.00000 0.9362 Residual 11 14.9568 1.3597 Total 12 234.4075 As with students in grades 2 and 6 there is a statistically significant negative linear relationship between age and achievement. The 2 R for grade 10 students (i.e., .94) is higher than the 2 R for grade 2 and grade 6 students (i.e., .89). As students get older the strength of the negative linear relationship between age and ac hievement for retained students increases. Figure 15 shows these same results. Fi g ure 15. SAT/9 mean total readin g NCE score re g ressed on a g e for retained students: Grade 10 STAR 2002 y = 1.0981x + 249.82 R2 = 0.9362 22 24 26 28 30 32 34 36 38 40 192193194195196197198199200201202203204205206 Age in MonthsSAT/9 Mean Reading NCE Score PAGE 23 Education Policy Analysis Archives Vol. 12 No. 49 23 Figure 15 shows that test scores decline for reta ined students and continue to do so through the age range. Figure 16 shows that as students get older th e difference in mean test score for the oldest and youngest students in the age normal peer group decreases. Fi g ure 16. Difference between mean total readin g NCE scores for the oldest and youngest students in the age normal peer group: STAR 2002 6.0 4.5 1.7 0 1 2 3 4 5 6 7 Grade 2Grade 6Grade 10 Grade LevelDifference in Mean Reading NCE Score Figure 16 shows that the variance in test scores for age normal peers decreases over time. By grade 10 the variance is so small that there is no practical difference in the test scores. Figure 17 shows the difference in mean test scor es for the oldest and youngest students for retained students. PAGE 24 Age and Achievement 24 Fi g ure 17. Difference between mean total readin g NCE score for the oldest and youngest students in the retained group: STAR 2002 13.2 14.1 12.1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Grade 2Grade 6Grade 10 Grade LevelDifference in Mean Reading NCE Score First, it can be seen that the difference between the oldest and youngest students is greater for retained students than for age normal peers. Second, as retained students get older the difference remains somewhat constant. For retained students the variance in test scores does not decrease over time. Another way to evaluate the advantage of retenti on is to compare test scores for birth dates close to the entrance cut off date. For example, students in 2nd grade with November 1984 birth dates are the youngest students. These are the students who are believed to be most at risk of academic failure by Uphoff & Gilmor e (1985). Students with November 1983 birth dates are the retained students who have been given time to mature in ways that lead to academic success according to Uphoff & Gilmor e (1985). Older students with November birth dates should demonstrate higher academic performance than younger students with November birth dates. Table 10 compares test scores for older and younger students with November, October, and September birth dates for three grade levels. PAGE 25 Education Policy Analysis Archives Vol. 12 No. 49 25 Table 10 Mean SAT/9 total reading NCE score for retained students compared to age normal students by birth month Grade 2 Nov93 49.6Oct93 48.3Sep93 46.4 Nov94 48.4Oct94 49.2Sep94 49.7 Difference 1.2 0.9 3.3 Grade 6 Nov89 48.8Oct89 47.1Sep89 45.5 Nov90 46.5Oct90 47.2Sep90 47.8 Difference 2.3 0.1 2.3 Grade 10 Nov85 38.0Oct85 37.0Sep85 37.6 Nov86 40.5Oct86 40.9Sep86 41.6 Difference 2.5 3.9 3.9 Retaining students with November birth dates does not on average provide a large academic advantage over younger classmates. In grade 2 retained students score on average 1 NCE point higher than their younger classmates and in grade 6 they score 2 points higher. By grade 10 older students score 2 and a half points lower than their younger classmates. Older students with October and September birth dates score lower than their younger classmates. These analyses continue to undermine the contention that retention provides students an academic advantage over their younger classmates. As stated, even though most students will be older than their age normal peers for three reasons: they started school late, spent two years in a transitional kindergarten or first grade program, or were flunked and forced to re peat a grade, there may be other reasons students are older than their classmates. Fo r example, maybe there are special education students who are older because they participa te in ungraded programs but are forced to indicate a grade level for testing purposes. Mayb e there are English learners (EL) who enter the system older than their classmates. Maybe sp ecial education and EL students artificially depress the test scores of older students. To examine this possibility, special education and EL students were removed from the analysis. Fi gure 18 shows average reading test scores by age in months for 2nd graders with special education an d EL students were removed from the analysis. PAGE 26 Age and Achievement 26 Fi g ure 18. Mean SAT/9 total readin g score b y a g e in months: Grade 2 STAR 2002: No special e ducation or EL students, n = 271,255 20 25 30 35 40 45 50 55 60 65 70 75Nov94 Oct94 Sep94 Aug94 Jul94 Jun94 May94 Apr94 Mar94 Feb94 Jan94 Dec93 Nov93 Oct93 Sep93 Aug93 Jul93 Jun93 May93 Apr93 Mar93 Feb93 Jan93 Dec92 Nov92858687888990919293949596979899100101102103104105106107108109 Birth Month & Year Age in MonthsSAT/9 Mean Reading NCE Score Overall test scores are higher when special education and EL students are not included. However, the pattern of scores by age in months is very similar to earlier analyses. That is, for age normal peers scores go up as age goes up. For retained students scores go down as age goes up. Removing special education and EL students does not modify earlier conclusions. The pattern of mean test scores by age migh t differ by subgroup. To evaluate gender differences, the average total reading NCE scores for students in grade 2 was calculated by age in months and by gender. Figure 19 shows these results. PAGE 27 Education Policy Analysis Archives Vol. 12 No. 49 27 Fi g ure 19. SAT/9 mean total readin g score b y g ender and a g e in months: Grade 2 STAR 2002: Female n = 222,969, male n = 232,439 20 25 30 35 40 45 50 55 60 65 70 75Nov94 Oct94 Sep94 Aug94 Jul94 Jun94 May94 Apr94 Mar94 Feb94 Jan94 Dec93 Nov93 Oct93 Sep93 Aug93 Jul93 Jun93 May93 Apr93 Mar93 Feb93 Jan93 Dec92 Nov92858687888990919293949596979899100101102103104105106107108109 Birth Month & Year Age in MonthsSAT/9 Mean Reading NCE Score Female Male The pattern of mean test scores by age is c onsistent for females and males. There are no gender differences. For both females and male s there is a positive relationship between age and achievement for the age normal peers. For students who have been retained there is a negative relationship between age and achievement. To evaluate SES differences, the average total reading NCE scores for students in grade 2 was calculated by age in months and by th e national school lunch program (NSLP). NSLP indicates whether or not students receive free or reduced lunch. Participation in NSLP is an indicator of lower SES. No NSLP is an indi cator of higher SES. Figure 20 shows these results. PAGE 28 Age and Achievement 28 Fi g ure 20. SAT/9 mean total readin g score b y NSLP and a g e in months: Grade 2 STAR 2002: NSLP n = 263,290, no NSLP n = 192,348 20 25 30 35 40 45 50 55 60 65 70 75Nov94 Oct94 Sep94 Aug94 Jul94 Jun94 May94 Apr94 Mar94 Feb94 Jan94 Dec93 Nov93 Oct93 Sep93 Aug93 Jul93 Jun93 May93 Apr93 Mar93 Feb93 Jan93 Dec92 Nov92858687888990919293949596979899100101102103104105106107108109 Birth Month & Year Age in MonthsSAT/9 Mean Reading NCE Score NSLP No NSLP The pattern of mean test scores by age is somewhat different for NSLP and no NSLP students. No NSLP students have the familia r patter of a positive relationship between age and achievement for the age normal peers an d a negative relationship between age and achievement for retained students. Although NS LP students have lower average test scores, there is a positive relationship between age and achievement for the age normal peers. For retained students there is a negative relati onship between age and achievement as compared to age normal students. Average test scores for retained are lower than those for age normal students. However, as NSLP students get olde r their average test scores dont show the decline seen in other figures. Average test scor es tend to flatten out. It may be that the average test scores have gotten so low that it is difficult for average scores to get any lower. Entrance Age Patterns in California Public Schools Even though evidence from this study and othe rs indicates that being older on average does not provide an academic advantage, there may be interest in the propor tion of delayed entry students that make up the overage population. There may also be interest in the extent of PAGE 29 Education Policy Analysis Archives Vol. 12 No. 49 29 academic redshirting in California public schools. Available data does not distinguish students who started school late from those w ho were retained for other reasons but it may be possible to make some inferences by looking at age frequency distributions. Figure 21 shows the percent of students in grade 2 by age in months who were tested in spring 2002. Fi g ure 21. A g e of second g rade students on December 1, 2001 for the STAR 2002 test, n = 485,796 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109Age in MonthsPercent of Students Figure 21 shows that there is a smaller percent of students who are 85 months old relative to the other age normal months (i.e., 86 to 96 mont hs). There is a larger percentage of students with October birthdays (i.e., 86 months) but it is still less, relative to the other age normal months. The percentage of students with Sep tember birthdays increases but is still less than the other age normal months. Students with August birthdays (i.e., 88 months) are the first group of students to have a percent that is consistent with the other months in this age normal cohort. There is a smaller percent of st udents who are 85 to 87 months old relative to their age normal peers (i.e., 88 to 96 months) because some proportion of these late birthday students have experienced delayed school entry. PAGE 30 Age and Achievement 30 The ages of 97 months and higher represent the students who have been held back. The retained students who have experienced delayed school entry are most likely those closest to the entry age cut (i.e., 97 to 99 months). St udent demographic information does not contain information as to why students were held back. However, the number and proportion of redshirted students can be conservatively estimated. One way is to work with the age normal cohort (i.e., 85 to 96 months) and ignore ages 97 months and higher. First, find the average numb er of students who are in the age range from 88 to 96 months. For students in figure 21 this value is 35,899. Assume that this is the number of students who should be in each of the months 85 to 87. Next subtract the number of students who are actually in each of the months 85 to 87 from the average of 35,899 (i.e., 85 months: 35,899 27,049 = 8,850; 86 months: 35,8 99 30,298 = 5,601; 87 months: 35,899 33,410 = 2,489). Sum these va lues (i.e., 8,850 + 5,601 + 2,489 = 16,941). Finally, divide this sum by the total number of students who should be in the months 85 to 96 if there were no delayed school entry (i.e., % 9 3 792 430 941 16 )3. Four percent of the second grade cohort in figure 21 has been academically redshirted. This translates into 25 percent of the students with November birthdays (i.e., % 7 24 899 35 850 8 ), 16 percent of the students with October birthdays (i.e., % 6 15 899 35 601 5 ), and 7 percent (i.e., % 9 6 899 35 489 2 ) of the students with September birthdays. The estimated percents can be interpreted as probabilities. For example, age normal students with November birthdays have a 25 percent probability of experiencing delayed school entry. Age normal students with October birthdays have a 16 percent probability of experiencing delayed school entry. Another way to estimate the proportion of reds hirted students is to look at the whole grade level cohort. There are 485,796 students in this cohort and 71,945 of these second graders have been retained. The retained stud ents represent 15 percent of the total cohort (i.e., % 81 14 796 485 71,945 ). It is unlikely that more than te n to twelve percent of kindergarten and first grade students were flunked statewi de. That means the percent of students who experienced delayed entry would be around three to five percent. This is consistent with the first estimate4. Despite controversy in the research li terature about academic redshirting, a healthy number of students (i.e., 574 14 796 485 03 to 290 24 796 485 05 ) in California experience delayed school entry. In the second estimate it is unlikely that the th ree to five percent is spread evenly through the whole overage cohort. It is most likely con centrated around the th ree months closest to the entrance age cut (i.e., 97 to 99 months). These months represent 7.7 percent of the 3 The average value of 35,899 is sub stituted in the months 85 to 87. 4 There may be other ways to guesstimate th e percent of students who have been redshirted. The reader is invited to do so. PAGE 31 Education Policy Analysis Archives Vol. 12 No. 49 31 second grade cohort or 37,406 students. This means that delayed entry students make up 39 to 65 percent of the retained students closest to the cut (i.e., % 39 406 37 574 14 % 9 64 406 37 290 24 ). Given the proportion of students who are reds hirted, the drop in test scores for second grade students who are 97 to 99 months seems particularly shocking. Test scores begin to decline even though 39 to 65 percent of these students have experienced delayed school entry to provide them a cognitive and emotional advantage over their grade level peers. To evaluate whether data in figure 21 repr esent a recent phenomenon, the age frequency distribution for students in grade 2 who were tested in spring 1998 was calculated. Figure 22 shows these results. Fi g ure 22. A g e of second g rade students on December 1, 1997 for the STAR 1998 test, n = 445,7070% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109Age in MonthsPercent of Students PAGE 32 Age and Achievement 32 The pattern is similar to that of second grade students tested in who were tested in spring 2002. It provides some evidence that the practice of delayed entry for students whose birthday is close to the cutoff date has b een a practice in California for several years. To further evaluate whether data in figure 21 represent a recent phenomenon, the age frequency distribution was calculated for grade 11 students who were tested in spring 2002. Eleventh grade students would have been s econd graders during the 199293 school year. Figure 23 shows these results. Fi g ure 23. A g e of eleventh g rade students on December 1, 2000 for the STAR 2002 test, n = 377,4340% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217Age in MonthsPercent of Students The pattern is similar to that of students tested in second grade. This is additional evidence that the practice of delayed entry for students whose birthday is close to the cutoff date has been a practice in California for several years. To evaluate gender differences in delayed school entry, the percent of students in grade 2 was calculated by age in months and by gender. Figure 24 shows these results. PAGE 33 Education Policy Analysis Archives Vol. 12 No. 49 33 Fi g ure 24. A g e of second g rade students b y g ender on December 1, 2001 for the STAR 2002 test, n = 485,5560% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109Age in MonthsPercent of Students Female Male The research literature indicates that males are more likely to be academically redshirted than females. Using estimation procedure #1, approximately 5 percent of the males and 3 percent of the females experienced delayed school entry. For males this represents 31 percent of the students with November birthday s, 22 percent of the students with October birthdays, and 11 percent of the students with September birthdays. For females this is 18 percent of the students with November birthd ays, 9 percent of the students with October birthdays, and 3 percent of the students with September birthdays. These data are consistent with the research literature. For example, ag e normal males with November birthdays have a 31 percent probability of experiencing delayed school entry while age normal females have a 22 percent probability. The research literature also indicates that parents identified as higher SES are more likely to delay school entry than those parents identified as lower SES. Figure 25 shows the percent of students in grade 2 by age in months by NSLP. PAGE 34 Age and Achievement 34 Using estimation procedure #1, these data indica te that approximately 5 percent of the no NSLP and 3 percent of the NSLP students ex perienced delayed school entry. For no NSLP students this represents 36 percent of the stud ents with November birthdays, 24 percent of the students with October birthdays, and 12 percent of the students with September birthdays. For NSLP students this is 16 percen t of the students with November birthdays, 9 percent of the students with October birthdays, and 3 percent of the students with September birthdays. These data are consistent wi th the research literature. Students who do not participate in NSLP (i.e., higher SES) ar e more likely to experience delayed school entry than students who do participate (i.e., lower SES). Parent education is another prox y for SES. Figure 26 shows the percent of students in grade 2 by age in months by parent education. Pare nt education in this case is represented by extremes. Low education means that the parent with the highest level of education in the household is a high school dropout. This is an indicator of lower SES. High education Fi g ure 25. A g e of second g rade students b y NSLP on December 1, 2000 for the STAR 2002 test, n = 485,7960% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109Age in MonthsPercent of Student s NSLP No NSLP PAGE 35 Education Policy Analysis Archives Vol. 12 No. 49 35 means that the parent with the highest leve l of education in the household has a college degree and has some post college education. This is an indicator of higher SES. These data indicate that approximately 6 percent of the students with the college + parent and 2 percent of the students with the hi gh school dropout parent experienced delayed school entry. For students with the college + parent this represents 37 percent of the students with November birthdays, 26 percent of the students with October birthdays, and 12 percent of the students with September birthdays. For the students with the high school dropout parent this represents 16 percent of the students with November birthdays, 7 percent of the students with October birthdays, and 2 percent of the students with September birthdays. Again, these data are cons istent with previous research and the data in figure 25. Students with a parent who has a college degree and some post college education (i.e., higher SES) are more likely to experience delayed entry than students whose most educated parent is a high school drop out (i.e., lower SES). Fi g ure 26. A g e of second g rade students b y parent education on December 1, 2000 for the STAR 2002 test, n = 166,0780% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109Age in MonthsPercent of Students High School Dropout College + PAGE 36 Age and Achievement 36 Discussion Findings indicate that for students in elementary school on average there is positive linear relationship between age and achievement for ag e normal peers. That is, on average, older age normal peers perform better academically th an their younger classmates. Even though there is positive linear relationship, the differe nce in average test scores between the oldest and youngest students is not great. The differe nce in average test scores between the oldest and youngest students gets smaller as grade le vel increases. By the time students reach 10th grade this difference is negligible. By 10th grade the positive linear relationship between age and achievement has disappeared. There is no academic advantage to being older, even for age normal peers, by the time students reach high school. For overage students there is on average a negative linear relationship between age and achievement for all grade levels. On average, overage students do less well than their classmates and the older they get the less we ll they perform. The negative relationship between age and achievement remains constant over time. The relationship between age and achievement is not content dependent. The relationship is consistent across reading and mathematics test scores. The relationship between age and achievement is also consistent across gender and SES differences. Even though on average there is a positive re lationship between age and achievement for age normal peers and a negative relationship for retained students, there is large variability in the individual test scores. Making strong inferen ces about student academic performance if age is known is not sound. That is, many of the youngest age normal students perform high and many of the oldest age norm al students perform low. The variability of individual test scores should not be interpreted to mean that it makes no difference as to whether or not students are retained. On average, retained students score lower than age normal peers. Since there are no valid or reliable instruments to identify whom would benefit from retention, retention of any kind is not a sound remediation strategy. The results also indicate that academic redshirting is practiced and has been in practice in California for several years. This is especially true for students with November birthdays. Consistent with previous research males are more likely than females to experience delayed school entry. Also, students from higher SES households are more likely to experience delayed entry than students from lower SES households. Even though research evidence does not suppor t delayed school entry, a certain percentage of parents have decided that it is in their childrens best interest to hold them back. Whether delayed entry is due to parental beliefs about the increasing curricular demands of kindergarten and first grade or other reasons, the pattern of behavior seems clear. Contrary to parental beliefs and educational poli cy, the results from this study argue against modifying entrance age policies, delaying school entry, implementing transitional kindergarten or first grade programs, or retaining students to improve educational PAGE 37 Education Policy Analysis Archives Vol. 12 No. 49 37 achievement. Proponents of these practices argue that they will improve educational achievement. These results do not support that argument. In addition, although it may be difficult to argue that being overage causes lower achievement, policies and practices that make students older than their age normal peers seem to inversely affect their educational ac hievement. When students are one year older than their classmates, their average academic performance declines and continues to decline the older they get. Maybe making students differe nt (i.e., older than their grade level peers) lowers their motivation to achieve. The research literature suggests that older students also are more likely to drop out of school. Learning exists along a continuum. With any group of people and with any content area people will learn at greater and lesser degrees. This occurs for a number of reasons. Three of the most obvious are ability, motivation and opportunity. Ability, motivation, and opportunity vary with individuals and thus learning also varies. In public education lip service is given to the notion that all students are expected to achieve at a certain level at every grade level. Even when grade level expectations are defined, students will master content at diffe rent rates. It makes no difference whether there are legislative decrees that include positive and/or negative consequences. There will be differences in student achievement. Student achievement exists along a continuum and making determinations about what it means to achieve mastery or grade level standards is arbitrary. Emrick (1971) stated: It is not difficult to show that traditional measurement procedures are inadequate, or at best arbitrary, as a method of identify student skill mastery. Glass (1978) went even further when he wrote: To my knowledge, every attempt to derive a criterion score is either blatantly arbitrary of derives from a set of arbitrary premises. Delaying school entry or retaining students in other ways to ensure some arbitrary level of achievement is a futile exercise. It cannot be over emphasized that attaining a certain test score is not the same thing as achieving mastery, even if mastery could be defined. At best, schools can identify where students are and move them further along the continuum. Some students can achieve at greater levels if given additional instruction and time. This is different than the notion of mastery. It simp ly means that movement along the achievement continuum can be accelerated with additional support. However, if additional time means making students older than their classmates by more than a year, the additional time begins to have a negative effect. PAGE 38 Age and Achievement 38 References Abidin, R.R., Golladay, W.M., Howerton, A.L. (1971). Elementary school retention: an unjustifiable, discriminatory, and noxious educational policy. Journal of School Psychology 9 (4) 410417. Angrist, J.D. & Krueger, A.B. (1992). The effect of age at school entry on educational attainment: An application of instrumental variables with moments from two samples. Journal of the American Statistical Association 87 (418), 328336. Bellisimo, Y., Sacks, C.H. & Mergendoller, J.R. (1995). Changes over time in kindergarten holding out: Parent and school contexts. Early Childhood Research Quarterly 10 (2), 205222. Brent, D., May, D.C., & Kundert, D.K. (1996) The incidence of delayed school entry: A twelveyear review. Early Education and Development 7 (2), 1996. Bracy, G.W. (1989). Age and achievement. Phi Delta Kappan 70 (9), 732. Byrd, R.S., Weitzman, M.L. (1994). Predictors of early grade retention among children in the United States. Pediatrics 93 (3) 481487. Byrnes, D.A. (1989). Attitudes of students, pare nts, and educators toward repeating a grade. In L.A. Shepard & M.L. Smith (Eds.) Flunking grades (pp. 3463). London: Falmer. Carlson, E., Egeland, B., Jimerson, S., Rotert, M., & Sroufr, L. A. (1997). A prospective, longitudinal study of the correlates and consequences of early grade retention. Journal of School Psychology 35 (1), 325. Combs, F.E. & Tanner, C.K. (1993). Student r etention policy: The gap between research and practice. Journal of Research in Childhood Education 8 (1), 6977. Crosser, S.L. (1991). Summer birth date child ren: Kindergarten entrance age and academic achievement. Journal of Educational Research 84 (3), 140146. DeMeis, J.L. & Stearns, E.S. (1992). Rela tionship of school entrance age to academic achievement. Journal of Educational Research 86 (1), 2027. Dietz, C. & Wilson, B.J. (1985). Beginning school age and academic achievement. Psychology in the Schools 22 (1), 9394. Ellwein, M.C., Walsh D.J., Eads, G.M., & Miller, A. (1991). Using readiness tests to route kindergarten students: The snarled intersecti on of psychometrics, policy, and practice. Educational Evaluation and Policy Analysis 13, 159175. Emrick, J.A. (1971). An eval uation for mastery testing. Journal of Educational Measurement 8, 321326. PAGE 39 Education Policy Analysis Archives Vol. 12 No. 49 39 Glass, G.V. (1978). Standards and criteria. Journal of Educational Measurement 15 (4), 237261. Graue, M.E. & DiPerna, J. (2000). Redshirt ing and early retention: Who gets the gift of time and what are its outcomes? American Educational Research Journal 37 (2), 509534. Grissom, J.B. (1988). Structural equation modeling of retention and overage effects on dropping out of school Unpublished doctoral dissertation, University of Colorado, Boulder, CO. Grissom, J.B. & Shepard, L.A. (1989). Re peating and dropping out of school. In L.A. Shepard & M.L. Smith (Eds.) Flunking grades (pp. 3463). London: Falmer. Holmes, C.T. & Matthews, K.M. (1984). Th e effects of nonpromotion on elementary and junior high school pupils. Review of Educational Research 54, 225236. Holmes, C.T. (1989). Grade level retention eff ects: A metaanalysis of research studies. In L.A. Shepard & M.L. Smith (Eds.) Flunking grades (pp. 1633). London: Falmer. Kinard, E.M. & Reinherz, H. (1986). Birth date effects on school performance and adjustment: A longitudinal study. Journal of Educational Research 79 (6), 366372. Langer, P., Kalk, J.M., & Searls, D.T. (1984). Age of admission and trends in achievement: A comparison of blacks and Caucasians. American Educational Research Journal 21 (1), 6178. La Paro, K.M. & Pianta R.C. (2000). Predi cting childrens competence in the early school years: A metaanalytic review. Review of Educational Research 70 (4), 443484. May, D.C. & Kundert, D.K. (1995). Does delayed school entry reduce later grade retentions and use of special education services? Remedial and Special Education 16 (5), 28894. Meisels, S.J. (1992). Doing harm by doing good: Iatrogenic effects of early childhood enrollment and promotion policies. Early Childhood Research Quarterly 7, 155174. Narahara, M. (1998). Kindergarten entrance ag e and academic achievement. 21 p. ED 421 218. Niklason, L.B. (1984). Nonpromotion: A pseu doscientific solution. Psychology in the Schools, 21, 485499. Shepard, L. & Smith, M.L. (1988). Escala ting academic demand in kindergarten: Counterproductive policies. Elementary School Journal 89 (2), 135145. Shepard, L. & Smith, M.L. (1985). Boulder valley kindergarten study: Retention practices and retention effects Boulder, CO, Boulder Valley Public Schools. Shepard, L. (1997). Children not ready to lear n? The invalidity of school readiness testing. Psychology in the Schools 34 (2), 8597. PAGE 40 Age and Achievement 40 Smith, M.L. & Shepard, L.A. (1987). What does nt work: Explaining policies of retention in the early grades. Phi Delta Kappan 6 (2), 129134. Smith, M.L. & Shepard, L.A. (1987). Teachers beliefs about retention. In L.A. Shepard & M.L. Smith (Eds.) Flunking grades (pp. 3463). London: Falmer. Uphoff, J.K. & Gilmore, J. (1985). Pupil age at school entrance How many are ready for success? Educational Leadership 43, 8690. Uphoff, J.K. (Eds.). (1995). Real facts from real schools Rosemont, NJ: Modern Learning Press. About the Author James B. Grissom Standards and Assessment Division California Department Education 1430 N Street Sacramento CA 95814 (916) 3190361 Email: jgrissom@cde.ca.gov B.A. California State University, San Bernardino, 1970 M.A. University of Colorado, Boulder, 1984 Ph.D. University of Colorado, Boulder, 1988 PAGE 41 Education Policy Analysis Archives Vol. 12 No. 49 41 Education Policy Analysis Archives http:// epaa.asu.edu Editor: Gene V Glass, Arizona State University Production Assistant: Chris Mu rrell, Arizona State University General questions about appropriateness of topics or particular articles may be addressed to the Editor, Gene V Glass, glass@asu.edu or reach him at College of Education, Arizona State University, Tempe, AZ 852872411. The Commentary Editor is Casey D. Cobb: casey.cobb@uconn.edu. EPAA Editorial Board Michael W. Apple University of Wisconsin David C. Berliner Arizona State University Greg Camilli Rutgers University Linda DarlingHammond Stanford University Sherman Dorn University of South Florida Mark E. Fetler California Commission on Teacher Credentialing Gustavo E. Fischman Arizona State Univeristy Richard Garlikov Birmingham, Alabama Thomas F. Green Syracuse University Aimee Howley Ohio University Craig B. Howley Appalachia Educational Laboratory William Hunter University of Ontario Institute of Technology Patricia Fey Jarvis Seattle, Washington Daniel Kalls Ume University Benjamin Levin University of Manitoba Thomas MauhsPugh Green Mountain College Les McLean University of Toronto Heinrich Mintrop University of California, Berkeley Michele Moses Arizona State University Gary Orfield Harvard University Anthony G. Rud Jr. Purdue University Jay Paredes Scribner University of Missouri Michael Scriven University of Auckland Lorrie A. Shepard University of Colorado, Boulder Robert E. Stake University of IllinoisUC Kevin Welner University of Colorado, Boulder Terrence G. Wiley Arizona State University John Willinsky University of British Columbia PAGE 42 Age and Achievement 42 AAPE Editorial Board Associate Editors Gustavo E. Fischman & Pablo Gentili Arizona State University & Universida de do Estado do Rio de Janeiro Founding Associate Editor for Spanish Language (19982003) Roberto Rodrguez Gmez Hugo Aboites Universidad Autnoma MetropolitanaXochimilco Adrin Acosta Universidad de Guadalajara Mxico Claudio Almonacid Avila Universidad Metropolitana de Ciencias de la Educacin, Chile Dalila Andrade de Oliveira Universidade Federal de Minas Gerais, Belo Horizonte, Brasil Alejandra Birgin Ministerio de Educacin, Argentina Teresa Bracho Centro de Investigacin y Docencia EconmicaCIDE Alejandro Canales Universidad Nacional Autnoma de Mxico Ursula Casanova Arizona State University, Tempe, Arizona Sigfredo Chiroque Instituto de Pedagoga Popular, Per Erwin Epstein Loyola University, Chicago, Illinois Mariano Fernndez Enguita Universidad de Salamanca. Espaa Gaudncio Frigotto Universidade Estadual do Rio de Janeiro, Brasil Rollin Kent Universidad Autnoma de Puebla. Puebla, Mxico Walter Kohan Universidade Estadual do Rio de Janeiro, Brasil Roberto Leher Universidade Estadual do Rio de Janeiro, Brasil Daniel C. Levy University at Albany, SUNY, Albany, New York Nilma Limo Gomes Universidade Federal de Minas Gerais, Belo Horizonte Pia Lindquist Wong California State University, Sacramento, California Mara Loreto Egaa Programa Interdisciplinario de Investigacin en Educacin, Chile Mariano Narodowski Universidad Torcuato Di Tella, Argentina Iolanda de Oliveira Universidade Federal Fluminense, Brasil Grover Pango Foro Latinoamericano de Polticas Educativas, Per Vanilda Paiva Universidade Estadual do Rio de Janeiro, Brasil Miguel Pereira Catedratico Universidad de Granada, Espaa Angel Ignacio Prez Gmez Universidad de Mlaga Mnica Pini Universidad Nacional de San Martin, Argentina Romualdo Portella do Oliveira Universidade de So Paulo Diana Rhoten Social Science Research Council, New York, New York Jos Gimeno Sacristn Universidad de Valencia, Espaa Daniel Schugurensky Ontario Institute for Studies in Education, Canada Susan Street Centro de Investigaciones y Estudios Superiores en Antropologia Social Occidente, Guadalajara, Mxico Nelly P. Stromquist University of Southern California, Los Angeles, California Daniel Suarez Laboratorio de Politicas PublicasUniversidad de Buenos Aires, Argentina Antonio Teodoro Universidade Lusfona Lisboa, Carlos A. Torres University of California, Los Angeles Jurjo Torres Santom Universidad de la Corua, Espaa Lilian do Valle Universidade Estadual do Rio de Janeiro, Brasil xml version 1.0 encoding UTF8 standalone no record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchemainstance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd leader nam a22 u 4500 controlfield tag 008 c20049999azu 000 0 eng d datafield ind1 8 ind2 024 subfield code a E1100398 0 245 Educational policy analysis archives. n Vol. 12, no. 49 (September 18, 2004). 260 Tempe, Ariz. : b Arizona State University ; Tampa, Fla. : University of South Florida. c September 18, 2004 505 Age and achievement / James B. Grissom. 650 Education x Research v Periodicals. 2 710 Arizona State University. University of South Florida. 1 773 t Education Policy Analysis Archives (EPAA) 4 856 u http://digital.lib.usf.edu/?e11.398 xml version 1.0 encoding UTF8 standalone no mods:mods xmlns:mods http:www.loc.govmodsv3 xmlns:xsi http:www.w3.org2001XMLSchemainstance xsi:schemaLocation http:www.loc.govmodsv3mods31.xsd mods:relatedItem type host mods:identifier issn 10682341mods:part mods:detail volume mods:number 12issue 49series Year mods:caption 20042004Month September9Day 1818mods:originInfo mods:dateIssued iso8601 20040918 